Project Allocation and Scheduling System

FIND A SOLUTION AT Academic Writers Bay

CODD01 – A Project Allocation and Scheduling SystemAcademic: Coddington, AlexThe aim of this project is to implement a system which allows final year or MSc projects to beallocated to students. Below is a list of suggested requirements.– Lecturers and students should be able to enter, edit and view their project suggestions (e.g.students should be able to suggest their own projects).– Lecturers and students should be able to see a list of all project suggestions.– Students should be able to create a shortlist of their preferred projects.– Students should be able to submit their final project choices.– The system should be able to automatically allocate projects to students.– The allocation process (see above) should take into account the work load of project supervisors,e.g. some supervisors might only be able to supervise 2 students because of other commitmentswhilst others might be allocated a larger number of students.– The final allocation should be viewable by both students and staff.– A user should be able to automatically create a schedule for the project progress presentation ordemonstration day.The challenge for this project is researching suitable algorithms that automatically generate aproject for a reasonably large number of students (e.g. 100).CODD03 – Locating DefibrillatorsAcademic: Coddington, AlexThe aim of this project is to build a mobile app, and any background infrastructure required, thatwill help people to find the nearest defibrillator in the event of an emergency (this should take intoaccount indoor mapping as well as outdoors). The app should also locate health professionals thatare in the vicinity so that they could help anyone who becomes unwell.CODD04 – Monitoring Student PerformanceAcademic: Coddington, AlexProject is suitable for 2 studentsThis project involves designing and building an app which will allow Advisers of studies to monitorstudents’ performance during the academic year. The app should allow staff to record/viewattendance at labs/lectures/tutorials, coursework marks and feedback, confidential information suchas personal difficulties/medical issues.The app should be able to rate each student’s performance using attendance data and marks for eachclass. The rating could be colour-coded (e.g. green for greater than 80% attendance and greater than60% for coursework completed so far, black for less than 20% attendance and less than 20% forcoursework completed so far) with a text description. The rating should be communicated to eachstudent at periodic intervals.The app should both alert the class coordinator/class lecturer if a student’s performance is less thansatisfactory for that class and warn the student.The app should alert the year adviser of study and possibly the Head of Undergraduate Teaching ifthe student’s performance is less than satisfactory for more than one class. The app should also beable to give each student customised advice depending upon the confidential information disclosedby the student to their adviser. For example, if a student has been suffering from prolonged periodsof illness or has suffered a bereavement which is affecting their studies, the app should provide thestudent with information regarding Voluntary Suspension, the University Disability and Wellbeingservice, etc.The app should be designed to be as flexible as possible – e.g it could be able to interface withdifferent means of storing data (e.g. different database systems) and should allow users to view theinformation satisfactorily using different devices.CODD06 – Monitoring Work-Based LearningAcademic: Coddington, AlexFoundation apprentices, modern apprentices and graduate apprentices need to undertake workbased learning as part of their programme of study. This involves building an e-portfolio ofevidence to demonstrate that they meet various professionally-defined skills or competencies (e.g.British Computer Society Register of IT Technician or Chartered IT Professional competencies, orcompetencies specified by the SFIA, Skills Framework for the Information Age).In order to determine which skills or competencies to focus on, apprentices need to undertake a selfassessment in order to understand their strengths and weaknesses, and then create and update apersonal development plan which focuses on competencies that need further work.Work-based learning advisers provide guidance to apprentices to help them to identify which oftheir work-based tasks can be used as evidence that can be included in their portfolio. Advisers meetregularly with apprentices to advise them and also to record: a) whether or not their work-basedlearning progress is satisfactory; b) whether the evidence they submit as part of their e-portfolio issufficient to demonstrate that the associated competency has been met.This project involves building an app that allows work-based learning advisers to input informationregarding the various competencies and also to record notes of the meetings they have withapprentices. IT should also allow work-based learning advisers to generate feedback (e.g. using atraffic light system where red indicates the apprentice is making very poor progress). The appshould also allow apprentices to submit evidence against the various competencies and record notes(these could involve identifying which competencies to target in the near future, or recording adiary of the work carried out).CODD07 – Building a Job Review AppAcademic: Coddington, AlexMany companies conduct annual reviews with employees to discuss their achievements during thepast year, and to identify targets and development opportunities for the following year.There are many Employee Performance Management systems on the market, but most suffer fromthe inability to deliver a truly continuous feedback process for employees, and from the lack ofgenuine and meaningful 360 degree input from a range of people who have been close to theemployees’ work.The objectives of this project are to build a continuous feedback system which will enableemployees to easily record their day to day achievements against competencies “as they go”, forimmediate notification to all reviewers. This could be achieved via an alert mechanism that issupported by mobile phone native app notifications and dashboard reporting of achievements to bescored. The system should allow peer groups, clients and other 3rd parties, to record their feedback– this could be achieved by the introduction of a social ratings system. Peers and the client can oftenbe closer to the employee’s performance than line and project managers, and the introduction of asocial “rating/comment/discussion” system that is integral to the process could provide betterevidence of the employee’s performance over the course of the year.CODD09 – Contract ManagementAcademic: Coddington, AlexOrganisations often get other companies or freelance workers to bid for contracts for various typesof work (e.g. building jobs, etc). The idea is to build an app that allows potential employers to placea description of the work that they would like carried out and allows workers or small companies toput in a bid for that work. The system could implement different types of auction protocols that canbe used to decide which bid to accept. The idea is that the company awarding the contract willprefer to minimise costs and time taken, while the workers placing bids will try to obtain as high aprice as they can.There are various dimensions along which auction protocols can vary:a) winner determination, i.e. who gets the work and what they bid (in first-price auctions the agentthat offers the lowest price gets the work;b) whether the bids made by agents are known to each other (open-cry or sealed-bid auctions);c) how bidding proceeds (e.g. in a single round, or where the price starts high and successive bidsare for increasingly lower amounts, or where the price starts high and the price decreases insuccessive rounds).The app should be able to implement different auction protocols such as English auctions (firstprice, open cry, ascending auctions) or Dutch auctions (open-cry descending auctions). The appshould ideally take account of the ways in which auction protocols might be susceptible to lying andcollusion by both bidders and employers.The app should be generic (i.e. be applicable to a number of specific businesses, be able to port todifferent database applications, as well as different devices, such as laptops and mobile phones). Afurther deliverable of the project is to demonstrate that the app can be tailored to at least twodifferent business applications.ReferencesAn Introduction to MultiAgent Systems, Michael Wooldridge, 2nd edition, John Wiley, 2009DONG01 – Compositional visual reasoning via neural networkAcademic: Dong, FengAI has made very impressive progresses in the last decade, attributing to many new approachesdeveloped under the framework of deep neural network (so called deep learning). The latestadvance in research has allowed AI to do language translation, answer questions and write text – toname but a few. However, majority of the existing work in AI do not explicitly exhibit a logicalbased reasoning ability. As a result, the outcomes from deep neural network lack explainability,which is regarded as a significant drawback of the current deep learning driven approaches.Some important progress has been made in visual reasoning, which is the ability to read a pictureand use acquired knowledge to draw novel inferences or answer new questions, e.g. “how manypeople in the photo?”, “What is the colour of the box next to the red ball?”. This is one of thefundamental building blocks of the intelligent mind. However, such research is only at thebeginning stage. The pros and cons of the existing approaches need further evaluations andanalyses.This project is set to evaluate the existing approaches and analyse their strength and weaknesses. Itwill apply the existing models to the CLEVR tasks of visual question answering (VQA). VQA is achallenging task that requires responding to natural language questions about images. The CLEVRbenchmark was created to address this problem as the dataset features unbiased, highlycompositional questions that require an array of challenging reasoning skills, such as transitive andlogical relations, counting and comparisons, without allowing any shortcuts around such reasoning.The research will involve the latest research in deep learning, involving both the convolutional andrecurrent neural network . The student is expected to have background knowledge and practicalskills (e.g. python and pyTorch) in AI, especially in deep learning and neural network. You will beexpected to run models from the repository and carry out an evaluation with selected datasets.[1] Compositional Attention Networks for Machine Reasoning https://arxiv.org/abs/1803.03067[2] https://github.com/stanfordnlp/mac-network[3] Neural Discrete Representation Learning https://arxiv.org/abs/1711.00937[4] https://github.com/1Konny/VQ-VAEDONG02 – A case study of neural meta learningAcademic: Dong, FengTraditionally, a neural network is trained by learning its weights using backpropagation. In such anapproach, humans need to manually decide the architecture of the neural network, as well as toexperiment with different settings of hyperparameters, which is often labour intensive.Recently, automatic search for hyper parameters and neural network architectures have become hotresearch topics. The former is associated with hyper-parameter learning and meta learning, which isto design an AI model to learn “how to learn faster” across different tasks, and the latter is designedto discover new neural network architectures by employing different search strategies.This project is set to evaluate the latest research advance in these areas through literature reviewsand through experiments with these models using benchmarks to report their performance. This willbe based on the open datasets and models that are publicly available.The student is expected to have background knowledge and practical skills in AI, especially in deeplearning and neural network. You will be expected to run the models from the repository and carryout an evaluation with selected datasets.[1]Bilevel Programming for Hyperparameter Optimization and Meta-Learning,https://arxiv.org/abs/1806.04910[2] https://github.com/prolearner/hypertorch[3] Generating Neural Networks With Neural Networks, https://arxiv.org/abs/1801.01952[4] https://hypnettorch.readthedocs.io/en/latest/DONG03 – An Evaluation of Diffusion Models vs Generative AdversarialNetworks (GANs) on Image SynthesisAcademic: Dong, FengThe latest progress in AI has greatly improved it’s ability in generating synthetic patient data thatare non-distinguishable from real data. More specially, the generative adversarial networks (GAN)can create synthetic data that look like real. This is achieved through measuring and minimisingdata distributions between real world and synthetic data, which have led to significant performancegain, especially in synthetic image generations.Until recently, GANs are regarded as the best generative model in image synthesis. However, somerecent work published recently has claimed that new diffusion based models can beat GANs inimage synthesis. This claim is only based on one research paper and is yet to be validated widely byother researchers in the AI research community.The goal of the project is to set to evaluate the outcomes from the diffusion model vs the GANmodels in terms of their capacity in generating synthetic images. This will be based on the opendatasets and models that are publicly available.The student is expected to have background knowledge in AI, data processing and practical skills(e.g. python and PyTorch). Knowledge about the generative adversarial networks and diffusionmodel is desirable but not essential.[1] Diffusion Models Beat GANs on Image Synthesis, https://arxiv.org/abs/2105.05233[2] https://github.com/openai/guided-diffusionDONG04 – Explainable Deep Learning ModelsAcademic: Dong, FengOne of the major limitations of deep neural networks is its poor explanability. Most of them act likea black-box. At the moment, it is generally recognised that model performance and explainabilityact as two requirements that are hard to be met simultaneously – often a better performed modelcomes with poor explainability. To break this deadlock, it is expected that future deep neuralnetworks will need to significantly improve their explanability.Recent research in deep learning have come up with new approaches that can explain the predictionoutcomes from deep neural networks to a certain extent. Examples include LIME, DeepLift – toname but a few. Especially, the recent work in SHAP (SHapley Additive exPlanations) has offeredan unified approach based on game theory to explain the output of any machine learning model.The goal of the project is to evaluate the outcomes from the SHAP in terms of their performanceand explanability in different use cases. This will be based on open datasets that are publiclyavailable. We will also use models from the GitHub repository as the starting point.The student is expected to have background knowledge in AI, data processing and practical skills(e.g. python and PyTorch), especially in deep learning and neural network. You will be expected torun the models from the repository and carry out an evaluation with selected datasets.[1] A Unified Approach to Interpreting Model Predictions https://arxiv.org/abs/1705.07874[2] https://github.com/slundberg/shapDONG05 – Fluid simulation vis deep learning approachesAcademic: Dong, FengFluid simulation is typically performed by solving a governing differential equation with numericalapproaches. While such traditional approaches have led to a lot of success in the past, they alsosuffer from limitations as some aspects of natural fluids are very difficult to be simulatednumerically. In recent years researchers have started to investigate new approaches by takingadvantage of deep neural networks. This has led to successful examples in a number of use cases.Deep fluids was proposed in recent years for fluid simulation in computer animation and computergraphics. It creates visually appealing results by using novel generative AI models together with theconvolutional neural network architecture to synthesise artificial fluid. It is reported that the methodcan model a wide variety of fluid behaviours, allowing for fast simulation on modest computers.The goal of this project is to set to evaluate the deep fluid model in terms of its performancethrough a number of use cases. This will be based on the open datasets that are publicly available.We will also use models in the GitHub repository as the starting point.The student is expected to have background knowledge in AI, data processing and practical skills(e.g. python, Tensorflow or PyTorch), especially in deep learning and neural network. You will beexpected to run the models from the repository and carry out an evaluation with selected datasets.[1] Deep Fluids: A Generative Network for Parameterized Fluid Simulations[2] https://github.com/byungsook/deep-fluidsDONG06 – Bayesian Neural NetworksAcademic: Dong, FengBayesian learning provides an important framework for machine learning method. Under thisframework, the parameters of the model do not just have single values learned through training.Instead, each parameter value is modelled with a distribution (e.g. normal distribution). A machinelearning model that is equipped with such parameter distributions can offer better prediction bytaking into account all possible parameter values. This also supports uncertainty estimation toprovide us with the information about uncertainty in model prediction outcomes.While Bayesian learning has been well applied to many machine learning models, including linearregression model and polynomial (non-linear) models, the research of applying Bayesian learning toneural networks has yet been completed. While there is a good body of work already, moreevaluations need to be done to investigate and compare the performance of different approaches.The goal of this project is set to evaluate the Bayesian neural network in terms of their performanceunder the framework of a recent open source package. This will be based on the open datasets thatare publicly available. We will also use models from the GitHub repository as the starting point.The student is expected to have background knowledge in AI, data processing and practical skills(e.g. python, Tensorflow or PyTorch), especially in deep learning and neural network. You will beexpected to run the models from the repository and carry out an evaluation with selected datasets.[1] https://arxiv.org/abs/1506.02142[2] https://github.com/JavierAntoran/Bayesian-Neural-NetworksDONG07 – A retrospective study to evaluate a new Digital cytology system atNHS LanarkshireAcademic: Dong, FengThe recently conversion of the primary test for cervical screening from cervical cytology to an HPVtest has been approved by the Scottish Government. The business case included procurement of theHPV test and the cytology triage test. The contract was awarded to a company to work closely withthe laboratories to introduce the HPV test. The Hologic imager system linked to microscope basedreview stations have been in place in the Scottish cervical screening laboratories since 2011 andhave increased the sensitivity of the test and reduced false negative results.This also included the option to adopt a new Digital cytology system that was still in development.The new technology uses artificial intelligence and image analysis algorithm and high resolutionmonitors to identify and present a gallery of 30-60 images on a monitor of the areas which are mostlikely to contain abnormal cells rather than via a microscope. The screener will still be required toidentify any abnormal cells within the gallery but will no longer be required to screen the slide andlocate potentially abnormal cells. This approach has the potential to reduce variability and improvesensitivity and efficiency by focusing the screener on the gallery.This project is set to collaborate with NHS Lanarkshire on an appropriate design of the research tovalidate the new Digital cytology system that uses machine learning / AI to pre-screen cytologyimage with retrospective data. There is access available to historical data. We will need to involvemultiple operators / screeners / etc., to assess reliability / inter-operator variability, etc at thehealthcare service.The student is expected to have background knowledge in health data analysis and practical skills,especially in the evaluation of machine learning and AI models (e.g. precision, sensitivity andspecificity) .DONG08 – Neural networks for natural language processing and programminglanguage processingAcademic: Dong, FengCan you imagine that a computer can automatically write software for you if you tell the computerwhat you need using natural language? Also, do you want the computer to automatically generatedocumentations from your source code to save your time? This project is set to focus on a blue-skyresearch that allows this to happen.More specifically, the project is set to evaluate the performance of deep neural networks inprogramming language and natural language, and investigate approaches to convert between thesetwo languages. New AI approaches will be studied to enable automated generation of computerprograms in source code from natural languages and vice versa.English will be the selected natural language, and we will not be restricted to using a specificprogramming language (i.e. the student can select a targeted programming language to work on).AI has made very impressive progresses in the last decade, attributing to many new approachesdeveloped under the framework of deep neural network (so called deep learning). The latestadvance in research has allowed AI to do language translation, answer questions and write text – toname but a few. While computer source code bears a lot of similarities to natural languages, it hasmany unique features. Some progresses have been made to write source code by AI, for example,by using graph neural network. However, this is still a new area and hence new studies are stillneeded.The research will involve the latest research in recurrent neural network, including the use oftransformers, BERT and graph neural network.The student is expected to have background knowledge and practical skills in AI, especially in deeplearning and neural network for natural language processing. Knowledge about the latestdevelopment in AI for computer source code are desirable.DONG09 – Evaluation of deep reinforcement learning and its applicationthrough a case study in computer gamesAcademic: Dong, FengDeep reinforcement learning has significantly improved the performance of traditionalreinforcement learning in recent years. The main initiative of reinforcement learning is to allow anAI agent to learn how to achieve goals in an environment through gaining rewards. By employingneural networks, the research community is able to overcome challenges that existed in manytraditional reinforcement learning approaches, such as Q-learning and policy gradient algorithms.This was especially demonstrated in successful computer game applications such as Atari, Go andRace Driving etc.However, despite these recent progresses, deep reinforcement learning still faces significantchallenges such as sparse rewards. Correspondingly, this project is set to evaluate the latestadvances in deep reinforcement learning, including imitation learning, behavior cloning, inversereinforcement learning, hierarchical reinforcement learning and so on, report their performance andanalyse their limitations.The study will be carried out in the context of designing and implementing a computer game. Thespecific format of the game can be decided by the student. The purpose of making the game is toprovide a context to study the performance of a variety of approaches in deep reinforcementlearning.The student is expected to have background knowledge and practical skills in AI, especially in deeplearning, neural network and reinforcement learningDONG10 – A case study of neural architecture search & hyper parametersAcademic: Dong, FengTraditionally, a neural network is trained by adjusting its weights using back propagation. In suchan approach, humans need to manually decide the architecture of the neural network, as well as toexperiment different settings of hyper parameters, which is often labour intensive.Recently, automatic search for hyper parameters and neural network architectures have become hotresearch topics. The former is named as “meta learning”, which is to design an AI model to learn“how to learn faster”, and the latter is designed to discover new neural network architectures byemploying different search strategies.This project is set to evaluate the latest research advance in these areas through literature reviewsand through experiments using benchmarks to report their performance. A case study will be carriedout in a selected application context (e.g. health). Potential approaches for further improvement willbe analyzed and discussed.The student is expected to have background knowledge and practical skills in AI, especially in deeplearning and neural network.GOOD02 – Automatically Generate Programming ExercisesAcademic: Goodfellow, MartinThe best way to learn programming is to practice writing code. The more practice the better. If aprogramming lecturer discovers that a class is struggling with a particular concept they will writemore coding examples and exercises to address this concept. However, it can be time consuming tocontinuously think up new examples and exercises. Therefore, it would be useful to write a tool thatcould automatically generate these exercises for basic programming challenges based on userrequirements.AimDevelop a web application capable of automatically generating programming problems for aspecific category or based on given user parameters.ChallengesDevelop a web application which can:– Provide a pre-existing library of programming exercises– Automatically generate programming exercises for a specific category, e.g., loops– Automatically generate programming exercises based on user parameters, e.g., must include a loopand a conditional.– Automatically generate Other Article: to exercise– Run code within the applicationGOOD03 – Visualising Code ExecutionAcademic: Goodfellow, MartinWhen learning to program students sometimes struggle to understand error messages and how to fixthem or for logical errors finding and understanding where they are going wrong. Tools likedebuggers can help with this process but for beginners these are sometimes hard to understand orfollow. Being able to visualise what is contained in variables and data structures throughout theexecution of a program can greatly help learners discover logical errors in their code. Therefore, theaim of this project is to provide a web application that allows users to step through code whilevisualising what is contained in the variables and data structures.AimDevelop a web application which visualises the contents of variables and data structures throughoutprogram executionChallengesDevelop a web application that which can:– Allow users to run code in the browser– Allow users to step through/into code– Visualise what is in data structure(s) during execution– Show what is in variables during execution– Help explain common searching and sorting algorithmsGOOD04 – Educational Data Analysis GameAcademic: Goodfellow, MartinGamification is the use of game elements or mechanics in non-game environments. Research hasshown it is an effective way to improve performance, motivation and engagement in education. Thegoal of this project is to create a web application game that teaches students about data analysis andgives them the skills to perform exploratory data analysis (initial investigation performed on adataset in order to understand it).GOOD05 – Educational SQL GameAcademic: Goodfellow, MartinGamification is the use of game elements or mechanics in non-game environments. Research hasshown it is an effective way to improve performance, motivation and engagement in education. Thegoal of this project is to create a web application game that teaches students how to use SQL.GOOD06 – Visualising Design PatternsAcademic: Goodfellow, MartinSoftware design patterns define a general reusable solution to problems which commonly occur insoftware design. This is an important topic which is taught within our software engineering classes.To aid student understanding of design patterns visualisations and code samples of different designpatterns are used. The aim of this project is to develop a tool to help with this.AimDevelop a web application that teaches users about design patterns through visualisations and codesamples.GOOD07 – Explainable AIAcademic: Goodfellow, MartinMachine Learning (ML) models are used in a huge number of software systems that can be used tomake potentially critical decisions for a business. However, a lot of these systems are black boxesand the users don’t understand how the models arrived at their decision. In this situation it can behard to trust that the AI is making a good decision, if we don’t understand the model’s reasoning.This has led to the area of Explainable Artificial Intelligence (XAI). XAI is AI where decisions canbe easily understood by humans. This makes it easier for humans and AI to work together byconfirming and challenging existing knowledge as well as being able to generate new assumptions.The aim of this project is to explore effective ways of teaching users how selected ML algorithmswork to aid in their understanding of decisions made by an ML model.HALV01 – University Information ServicesAcademic: Halvey, MartinProject is suitable for 3 studentsThese projects are offered in conjunction with Donna Brawley (Senior Applications Analyst,Business Information Services). The aim is to examine different potential new applications forstudents at Strathclyde University, in particular around the Strathclyde App. For IM students theprojects will involve requirements gathering through iterative design involving multiplestakeholders as a case study. For ACS students you must implement and test a prototype, IM canalso do this. Indicative topics are as follows:1.Timetabling (online v on campus challenges – mixed models of teaching logistics and handlingstudent enquiries)2. Room and Resource Booking on campus to facilitate blended learning and staff agile working3. Mobile App – planning for Version 4.0 – a complete new look and feel with user driven services4. Mobile ‘accessibility’ challenges5. Azure Authentication modelsHEDG01 – Numerical solution of compositional stochastic gamesAcademic: Hedges, JulesProject is suitable for 3 studentsCompositional game theory is a recent new approach to economic game theory inspired by modernprogramming language design (see https://arxiv.org/abs/1603.04641 for example). This projectexplores a new connection between compositional game theory, optimal control and reinforcementlearning.This project will involve re-implementing the basic mathematical framework of compositionalgame theory, currently implemented in Haskell, in a language such as Julia or Python that is suitablefor numerical computation while still being high level enough for the theoretical aspects of theframework. This implementation will then be used to demonstrate the numerical computation ofMarkov equilibria of stochastic games, using methods of dynamic programming. Time-permitting itmay also be used to demonstrate basic reinforcement learning algorithms such as Q-learning.This project would be suitable for anyone familiar with numerical programming or keen to learnabout it, and anyone interested in the mathematical / game-theoretic foundations of reinforcementlearning or optimal control. Familiarity with functional programming would also be helpful, inorder to understand the current implementation and the theory behind it.HEDG02 – Rendering string diagrams with Haskell’s Diagrams libraryAcademic: Hedges, JulesProject is suitable for 3 students“String diagrams” are a 2-dimensional notation similar to circuits that are used in a variety ofsubjects including quantum mechanics, linguistics, game theory and many others. Despite theirusefulness, computer support for string diagrams is quite limited outside of specific use-cases.Haskell’s Diagrams library (https://archives.haskell.org/projects.haskell.org/diagrams/) is a graphicslibrary that is built to support a functional programming style. In theory, it is possible using thislibrary to render a string diagram using a fold over an algebraic datatype that represents the“underlying” algebraic theory, known as monoidal categories (see https://arxiv.org/abs/1908.10660).The goal of this project is to make this a reality.This project would suit a mathematically-minded student interested in functional programming. Ifsuccessful, the result of this project may be the first part of a small ecosystem of tools for workingwith string diagrams in Haskell.HEDG03 – Graphical differentiable programming in Julia with CatlabAcademic: Hedges, JulesProject is suitable for 3 students“String diagrams” are a 2-dimensional notation similar to circuits that are used in a variety ofsubjects including quantum mechanics, linguistics, game theory and many others. One of their usesis to represent “computational graphs”, of which neural networks are an example.The Catlab library for Julia (https://algebraicjulia.github.io/Catlab.jl/latest/) is a library for workingwith string diagrams. The goal of this project is to extend Catlab with backpropagation forcomputational graphs. This will allow graphical differentiable programming (very much likeTensorFlow) in Julia. This will then be used to demonstrate visualising some simple machinelearning algorithms.Depending on the exact direction, this project would either suit a student interested in highperformance numerical computing and Julia, or a more theoretically-minded student interested inthe foundations of machine learning and differentiable programming.KUPK01 – Ontology reasoning algorithmAcademic: Kupke, ClemensDescription Logics allow to represent knowledge about specific domains of interest, e.g. one canformulate facts about the world such as “an employee is a person” and “every person has a name”.We then might ask the question whether a given statement is consistent with our generalbackground knowledge. For example the statement “there is an employee x and x has no name” willnot be consistent with the above mentioned facts, although without the above knowledge it mightbe.The way we check consistency of a statement is by checking whether there exists a model for it thatvalidates all background knowledge.This project is about implementing an algorithm that is capable of doing the above mentionedconsistency check. In addition, the implementation should either be extended to incorporateconstants (allowing statements of the type “Colin is a person.”) or more sophisticated conditionsinvolving counting or numbers such as “the majority of all actors is acting in at least 5 movies”.Requirements for the project are an interest in logic & algorithm design.KUPK09 – How to best solve parity gamesAcademic: Kupke, ClemensParity games are an important tool in formal verification that allow to reduce verification tasks tothe question of finding a winning strategy in a game. A variety of algorithms for solving paritygames has been proposed and implemented. This project is about the question how we can decidewhich algorithm one should pick for a given game. This should be done by using machine learningtechniques to create a model that is able to predict which algorithm is best suitable to solve anygiven game. The quality of the predictions should be analysed in detail.Publicly available collection of benchmark games:https://github.com/jkeiren/paritygame-generatorTool that implements the various algorithms for solving parity games:https://github.com/tcsprojects/pgsolver/KUPK10 – Property advisorAcademic: Kupke, ClemensThis aim of this project is to investigate how publicly available data on property prices (e.g. onzoopla) can be used to support people who are looking to buy a home. As part of the project a toolshould be created that allows to*) predict price development within an area based on recent sales*) find areas that might be on the cusp of becoming popular*) find areas that are “similar” to the current area of interest.The precise list of features will need further clarification and can be negotiated at the beginning ofthe project.KUPK11 – Twitter BotAcademic: Kupke, ClemensThe overall goal of the project is to research how twitter bot accounts can be recognised by theirtweet activity and profile. To his aim a tool should be created that is able to associate scores withany given twitter account. In addition, the tool should provide reasons for the score and someinformation on its reliability. For training the tool we can rely on data athttps://botometer.osome.iu.edu/bot-repository/index.htmlKUPK12 – Model Learning for Transition SystemsAcademic: Kupke, ClemensLabelled Transition Systems (LTSs) are mathematical models that allow to faithfully represent thebehaviour of a large variety of computational systems. To be able to verify properties of suchsystems it is vital to keep the size of the model as small as possible without losing essentialinformation.This project aims to implement a new, on-the-fly learning algorithm for LTSs that is based on modallogic. A first prototype that can learn the quotient of an LTS modulo trace equivalence has beenimplemented. This prototype needs to be extended in several ways, e.g. we would want to learn the(weak) bisimulation quotient of an LTS and/or weighted/probabilistic transition systems. Ideally thealgorithm should be implemented parametric in the type of system under consideration and in thesystem equivalence (trace equivalence, bisimulation equivalence etc.) Good languages for thisprojects are Haskell or Python.KUPK13 – Implicit Information on WikidataAcademic: Kupke, ClemensWikidata (https://www.wikidata.org/) is an open knowledge base that can be read and edited byusers worldwide. Knowledge is encoded in a graph-like way, the graph can be queried using theSPARQL query language. The goal of this project to see how implicit information can be deducedfrom the shape of the graph. For example, Wikidata contains information about writers in a country– is it possible to find the most popular writer in a country by analysing the data? Or the mostpopular politician? (these examples are arbitrary – some creativity as to the information that wewould like to deduce is welcome!) One deliverable of the project should be a tool that displaysinteresting pieces of information about countries that we manage to deduce from the data.LENN01 – Mobile app for Encouraging Physical Acvitity in people living withDiabetesAcademic: Lennon, MarilynThe management of Long Term Conditions costs the UK millions of pounds each year. With anincreasingly population, we can expect to live longer and many of us will do this while living with,and managing conditions such as Asthma, COPD or Diabetes.Technology can play a huge role in enabling us to self manage these conditions pro-actively whilewe are younger and also from the comfort of our own homes as we grow older independently.Smartphones can be used for example to track symptoms, receive health directed advice andmessages tailored to individuals and present information back to us in an engaging and useful waythat we can use to learn about our condition and our lifestyle and stay well. Increasing peopls dailyphysical activity is one ssuch area of research that deserves attention.AimThe aim of this project is to design, build and evaluate a mobile smartphone app (android) whichsupports people who want to self manage their diabetes through monitoring and increasing theirdaily physical activity.The solution should include functionality to:(a) track symptoms and lifestyle information such as diet, food intake, exercise, blood glucoselevels etc(b) view physical activity over time (daily, weekly, monthly) and provide visualisations that areuseful to the individual, their friends and family and their caregivers(c) allow users to share information with friends and family perhaps via social media or other meansand consider ways to allow users to tailor what they share and how they share itThe student will also design and conduct a study to evaluate the usability of the app.Pre-requisitesThe student must be able to (or willing to learn) programming in an Android environment.The student must be interested in investigating the HCI (usability and user experience) of the appand develop it iteratively with users in mindThe student must be willing to investigate the needs and requirements of people living withDiabetes.LENN02 – Design of User Interfaces for Clinical Decision SupportAcademic: Lennon, MarilynIt is often said that AI is the future of health – but what we do with the data we collect about peopleand their health is crucial – and interface design has a huge role to play in making technologies forhealth and wellness a success.This project will involve app or web development as a means to explore different userinterfaces(UIs) for presenting risk in useful and meaningful ways to CLINICIANS.You must be able to design and develop dynamic interactive webpages or smartphone applicationsand/or prototype novel designs and test them in a systematic manner with solid research methods.The secondary aim will be to expmpirically explore which interfaces are most effective and/orpreferred in helping people to make clinical decisions.LENN03 – The use of AI to understand the sentiment in health and care recordsAAcademic: Lennon, MarilynUnderstanding the sentiment in health records (written or electronic) is important to help carers andprofessionals to make decisions about a persons care (whether they needs treatment and what thattreatment should be).AI (artificial intelligence) methods and tools might be able to help us extract meaning and patternsfrom textual notes and records and enable us to better automate the recognition of key words orphrases so that this task is done with the assistance of computers to make the task both faster andalso more accurate.This project will involve the design, development, and analysis of algorithms and tools for:1 – anynonmysing health records pre analysis2 – detecting patterns, key words,o or sentiment in health record daa3 – building simple models to predict what actions should be taken based on the patterns found inthe datasets.Students must have a strong programming background and/or an interest in AI and anaysis.LENN04 – Digital Tools to support Medication ManagementAcademic: Lennon, MarilynPeople are living longer and as a result many of us will at one point in our lives have to managemultiple conditions and possibly multiple medications.Remember to take medications, and tracking intercations between drugs is complex and making thistask easier could increase medication compliance, reduce the costs associated with the mismanagement of medications and make the user experience better.This project will involve the design, development and evaluation of digital solutions for helpingpeople manage their medications.You must be interested in both design and development to take this project.LENN05 – Interactive website for Medication ManagementAcademic: Lennon, MarilynPeople are living longer and as a result often have multiple health and care needs or conditions andrequire support to manage staying well for longer at home and in the community.Pharmaceutical Care (the management of medicines) is something that is particularly challenging.Helping people to manage their own medicines can help increase the benefits of their treatmentsand reduce the costs associated with complications that arise when people do not use theirmedicines as advised. Healthcare professionals, such as pharmacists and general practitioners,might also benefit from systems that help them to monitor and or manage medicines for theirpatients more safely and effectively.This project will involve working with a client (a larger research consortium) to design and developa website which highlights/illustrates(i) the user/patient journey and challenges around daily medicine management and/or(ii) healthcare professionals’ and carers’ needs in relation to better pharmaceutical care and/or(iii) current and future technologies (digital and non-digital) for supporting pharmaceutical careThe student should have strong design and web development skills and be willing and able to workwith other academics and professionals to gather requirements and content for the website. Thewebsite should be interactive and so experience in, or willingness to learn back end technologiessuch as SQL and scripting such as Java and/or PHP are essential (it will not be a static website).LENN06 – DigiPROM: An app for supporting patient reported outcomesmeasurementAcademic: Lennon, MarilynHolistic needs assessment using patient reported outcome measures is advocated for optimal cancercare delivery. Whilst this approach to cancer care has been advocated, what is apparent is that theuse of PROMS in clinical practice is not routine with barriers cited such as lack time to carry outassessment and reliance on paper questionnaires that make is hard for the clinician to identify whatpatient needs/concerns should be prioritised. Developing an app for this purpose may overcomesome of these barriers as it would automate the application and provide scoring based on thepatients reports/AimTo design and build a holistic needs assessment app for use by cancer nurse specialists in thedelivery of supportive careThe app might include features such as:• Use of electronic patient reported outcome measures• Automated scoring for clinicians to identify what patient/needs concerns to prioritise• Guidance for the clinician on how to deal with these needs using best practice/evidenceLENN07 – Glasgow Smiles Better: A Mobile Phone app for capturing thehappiness of a city and it’s citizensAcademic: Lennon, MarilynIs smiling really contagious? Can you capture the wellness of a city and it’s citizen’s?This project will involve the design, development and testing of an smartphone App for (i)capturing the wellness/happiness of individual citizens and (ii) visualising patterns of happinessacross the city. The App will notifiy people (at random times of the day) and ask the recipient howhappy they are via a simple cartoon ‘smiley’ face. The App will collect other data such asdemographics, location data, weather data and time of day data and explore the best ways toaggregate and visualise this data back to the user themselves and possibly other stakeholdersinterested in the wellbeing across the city. This information will be collated to provide aggregateddata for the level of happiness across the city. The App will use personal and aggregated data to givefeedback to recipients, for example: This morning you are the happiest person on Byres Road; Youare always the happiest on Fridays. Follow-up notifications will record happiness so the App cantrack the impact of feedback on happiness.The project will involve both research and investigation and also app design and development (thiswill most likely be done in either android or using html5).LENN08 – PMA: A positive messages smartphone appAcademic: Lennon, MarilynBackgroundWell times and appropriate wellness messages can have a direct affect on our mood, our attitudesand expectations and possibly our overall health and wellbeing. Mobile apps could be used todeliver personalised messages to the user at the right time, at the right place in order to achieve this.Little is known however what these messages should be, and how you determine the ‘right time’ andthe ‘right place’.AimThe aim of this project is to design, implement and evaluate a mobile smartphone app that can tailormessages to a user (personalisation) and can use context and user preferences to determine when theright time and place to deliver these messages are. The project will also look at the best ways toevaluate the usability and user experience of the app and the delivery and scheduling of themessages.Challenges•Creation of a simple usable app that allows users to personalise how the receive messages•Investigation of the best ways to detect, sense and learn when and how a message should bedelivered•Exploring how you can evaluate whether the messages are received, attended to, and if they have apositive or negative affect on the user based on mode and time of deliveryLENN09 – Reducing sedentary behaviour using personalised activitynotificationsAcademic: Lennon, MarilynThere are many thousands of smartphone apps that aim to increase peoples’ physical activity andreduce their sedentary behaviour. Sensors on the phone can be used to detect activity (or inactivity)and also to notify the user periodically to move or be more active. Persuasive techniques can beused to motivate the user to change their behaviours gradually so that they rely less on thetechnology and gradually learn the more health behaviours. When and how to remind or notifypeople without the app becoming disruptive or annoying is a challenge. Keeping someone ‘playing’or engaging with these active lifestyle apps long enough to change their behaviour in a positive wayis also a challenge.This project will involve designing and building a prototype app that will monitor a users physicalactivity (or inactivity) and empirically explore how and when to notify the user that an action isrequired. The output will be a working prototype app and findings from a user trial on which styleor schedule of notifications work best in terms of performance (which ones people to attend tobetter) and preference (which ones users actually like the best).MAGU01 – See it, say it, do it – development for an app for health professionalsin practiceAcademic: Maguire, RomaProject is suitable for 2 students(FOR 2 STUDENTS) – The See It, Say It, Do It (initiative was launched on Friday 6th May 2016).See it, Say it, Do it – YouTube. At that time, the project’s main aim was to effect a change in cultureacross the organisation by empowering all staff, carers, patients, and their families by escalatingtheir concerns or providing a vehicle for acknowledging good practice. There is now a need to rerefresh the initiate and and refocus its use solely for staff. Patients and carers now have a welldefined route to comment on their care through care opinion, which was not as sophisticated in2016. This link will allow access to the care opinion website:https://www.careopinion.org.uk/info/care-opinion-scotlandPart of the Care Opinion approach is to allow patients and carers to have a website where:• people can share honest feedback easily and without fear• stories are directed to wherever they can help make a difference, and• everyone can see how and where services are listening and changing in responseIt is the responsibility of all healthcare staff to raise concerns around safety, undermining, bullying,harassment, housekeeping, uniform compliance, etc. However, there is no simple or easy route to doso. Using a web portal or apple/android applications (apps) would allow staff will feel empoweredand seek a resolution to their problem, building on the bullet points of the care opinion approachabove.It is thought that developing a web portal and apps may be a suitable project for an M.Sc student toundertake. Whereby a student could explore the development of the portal and/or app. It isenvisaged that staff views in the approach would be explored through staff focus groups formedfrom a cross-section of the workforce. It would also be useful to explore the most effective waystaff feedback is captured, such as prefilled drop-down lists, free text, dictation/voice recognition,etc.If possible, by exploring the use of artificial intelligence, it would be hoped to group issues intothemes to provide alerts to those identified to monitor the portals and apps, allowing appropriateaction to take place.This project could include 2 students – one student to develop the portal/app and the other to look atthe potential of AI to be used in the application.MAGU02 – Development of an app to support the management of chronicconditions in dogsAcademic: Maguire, RomaDogs are a ‘persons best friend’ but as they age they can develop chronic diseases such as arthritis,diabetes that can impact on their quality of life. Digital health apps may play a role in supportingthe management of conditions in dogs and providing owners with information on best to providecare. These apps can also enable the needs/behaviours of the pet to be monitored and for thisinformation to be directly sent to the vet – enable virtual monitoring and remote consultation.The project will therefore focus on the development of an app for dog owners to remotelymonitoring and support dogs with chronic disease and where appropriate send this information totheir veterinarian. The project requires a student with a good level of programming skills andknowledge of co-design methodologies. The project will be supervised with Professor RomaMaguire and may involve working with dog owners and vets where feasible and appropriateMAGU04 – The development of an app to manage loneliness in older peopleAcademic: Maguire, RomaLoneliness is increasing in our older population and has been defined as one of the new socialdeterminants of health. According to Age UK, more than 2 million people in England over the ageof 75 live alone, and more than a million older people say they go for over a month withoutspeaking to a friend, neighbour or family member.People can become socially isolated for a variety of reasons, such as getting older or weaker, nolonger being the hub of their family, leaving the workplace, the deaths of spouses and friends, orthrough disability or illness. Whatever the cause, it’s shockingly easy to be left feeling alone andvulnerable, which can lead to depression and a serious decline in physical health and wellbeing.Someone who’s lonely probably also finds it hard to reach out. There’s a stigma surroundingloneliness, and older people tend not to ask for help because they have too much pride.The aim of this project is to co-design, develop and evaluate an app developed to reduce lonelinessin older people. The student will require a very good level of programming skills and haveexperience of user co-design.MAGU05 – Development of an app for people experiencing chronic painAcademic: Maguire, RomaChronic or persistent pain is pain that carries on for longer than 12 weeks despite medication ortreatment. Most people get back to normal after pain following an injury or operation. Butsometimes the pain carries on for longer or comes on without any history of an injury or operation.Chronic pain affects 1 in 5 people in Scotland. It can affect all ages and all parts of the body. It isn’tpossible to tell in advance whose pain will become chronic. But we know that people are morelikely to develop chronic pain during or after times of stress or unhappiness. People can alsoexperience chronic pain even after usual medical tests don’t provide an answer.The experience of chronic pain can have a significant impact on the lives of the individual affected.They can experience sleep deprivation, depression, fatigue and several other symptoms. For peoplewith severe chronic pain it can impact on their ability to work and impact on relationships withfamilies and friends. The aim of this project is to co-design, develop and evaluate an app to supportpeople with chronic pain at home and to enable them to communicate with health professionals asand when required. The student undertaking this project needs a good level of programming skillsand knowledge of the co-design processMARD11 – The concept of efficiency for Energy SystemsAcademic: Mardare, RaduEnergy savings and the control of energy consumption are probably some of the most importantproblems in our society, having major impact on various levels. One can understand the energyconsumption pattern of a house, institution, town, etc., as a computational model where the energyflow is labels the systems transition and govern its behaviour.The aim of this project is to develop a model of energy (computational) systems, for instance byextending the concept of transition system or finite state automata to include information aboutenergy consumption. The focus will be on understanding and defining what does it mean for anenergy system to be “more efficient” than another one. “More efficient than” is an order relationthat will be motivated and verified against logical specifications that one needs to properly define.MARD12 – Modeling the “cheaper than” relation for Priced ComputationalSystemsAcademic: Mardare, RaduMany systems of interest in engineering involve price information that are vital in decision making.The price could represent production costs in manufacturing applications, energy consumption costsin energy systems, market trading results or even computational costs of processes. The behavior ofsuch a system will depend not only of its input, but also of the price evolution that the system keepscomputing while it evolves. Hence, understanding and predicting the behaviour of such a systemagainst the price evolution is essential.The aim of this project is to develop a model of priced computational systems, for instance byextending the concept of transition system or finite state automata to include price information. Thefocus will be on understanding and defining what does it mean for a priced system to be “cheaperthan” another one. “Cheaper than” is an order relation that will be motivated and verified againstlogical specifications that one needs to properly define.MARD13 – The equivalence of Epistemic SystemsAcademic: Mardare, RaduEpistemic systems are systems of agents witnessing a certain reality and computing knowledgeabout it. They are intensively used in modeling, for instance of security systems where one wants tounderstand and control the information accessed by certain agents active on a network. Due to itsrelevance in applications, the field of epistemic logic had a considerable evolution in the lastdecades.This project aims at developing a couple of fundamental concepts of equivalence that might berelevant for epistemic systems. For instance, (i) what does it means for two agents to haveequivalent knowledge, or (ii) what does it means for two societies of agents to have equivalentknowledge, or even (iii) what makes two epistemic systems equivalent. To argue for possibleanswers to these questions one can make use of fragments of epistemic logic.MARD14 – The equivalence of Resource Dependent SystemsAcademic: Mardare, RaduIn the practice of modelling we are often challenged by systems that depend of various types ofresources for a proper functioning, such as batteries or fuel. Once the resources are consumed, thesystem ceases to evolve. It is that important to understand their behaviour in relation to resourceconsumption.The aim of this project is to develop a model of resource-dependent (computational) systems, forinstance by extending the concept of transition system or finite state automata to includeinformation about resource consumption. The focus will be on understanding and defining theequivalence of two resource dependent systems from the point of view of the resource consumptionwhile executing similar tasks. This definition will be verified against logic-based queries that willproperly investigate the consumption patterns.MARD15 – “Faster than” relation on Time Dependent SystemsAcademic: Mardare, RaduTime is a fundamental resource in computation. Even more, in robotics or cyber-phisical systems, acomputational phenomenon might have temporal restrictions that will directly influence itsbehaviours and often schedules are required to handle them.The aim of this project is to develop an appropriate model of time-dependent (computational)systems, for instance by extending the concept of transition system or finite state automata toinclude temporal information and restrictions. The focus will be on defining what it means for asystem to be “faster than” another system, as for instance when one system executes any task fasteragainst any possible schedule. This order can be logically characterized and verified.MARD16 – Deciding the resource efficiency of systemsAcademic: Mardare, RaduIn the practice of modelling we are often challenged by systems that depend of various types ofresources for a proper functioning, such as batteries or fuel. Once the resources are consumed, thesystem ceases to evolve. It is that important to understand their behaviour in relation to resourceconsumption.The aim of this project is to develop a model of resource-dependent (computational) systems, forinstance by extending the concept of transition system or finite state automata to includeinformation about resource consumption. The focus will be on defining what it means for a systemto be more efficient in utilizing its resources than another one. This is an order relation that willallow us to classify system from the perspective of their efficiency. The correctness of thisdefinition will be verified against logic-based queries that will properly describe the consumptionpatterns.MARD17 – Equivalence of Energy SystemsAcademic: Mardare, RaduEnergy savings and the control of energy consumption are probably some of the most importantproblems in our society, having major impact on various levels. One can understand the energyconsumption pattern of a house, institution, town, etc., as a computational model where the energyflow is labels the systems transition and govern its behaviour.The aim of this project is to develop a model of energy (computational) systems, for instance byextending the concept of transition system or finite state automata to include information aboutenergy consumption. The focus will be on understanding and defining the equivalence of twoenergy systems from the point of view of energy consumption while executing similar tasks. Thisdefinition will be verified against logic-based queries that will properly investigate the consumptionpatterns.MARD18 – The analysis of Priced Computational SystemsAcademic: Mardare, RaduMany systems of interest in engineering involve price information that are vital in decision making.The price could represent production costs in manufacturing applications, energy consumption costsin energy systems, market trading results or even computational costs of processes. The behavior ofsuch a system will depend not only of its input, but also of the price evolution that the system keepscomputing while it evolves. Hence, understanding and predicting the behaviour of such a systemagainst the price evolution is essential.The aim of this project is to develop a model of priced computational systems, for instance byextending the concept of transition system or finite state automata to include price information. Thefocus will be in understanding and defining what does it mean for two priced systems to haveidentical behaviours, despite their structural differences. The price equivalence of systems will beverified against logical specifications that one needs to properly define.MARD19 – Simulating Epistemic SystemsAcademic: Mardare, RaduEpistemic systems are systems of agents witnessing a certain reality and computing knowledgeabout it. They are intensively used in modeling, for instance of security systems where one wants tounderstand and control the information accessed by certain agents active on a network. Due to itsrelevance in applications, the field of epistemic logic had a considerable evolution in the lastdecades.This project aims at developing concepts of simulation that can be useful in applications. Forinstance, what does it mean that an agent can simulate another agent? This can be a useful conceptin security, where dissimulated behaviours can be used to avoid security checks.MARD20 – Equivalent Time Dependent SystemsAcademic: Mardare, RaduTime is a fundamental resource in computation. Even more, in robotics or cyber-phisical systems, acomputational phenomenon might have temporal restrictions that will directly influence itsbehaviours and often schedules are required to handle them.The aim of this project is to develop an appropriate model of time-dependent (computational)systems, for instance by extending the concept of transition system or finite state automata toinclude temporal information and restrictions. The focus will be on defining what it means for twosystems to be temporal equivalent. Such a concept could be, for instance, defined so that twoequivalent systems behave in the same way against any conceivable schedule.MCCA08 – Connecting children during periods of isolation: Developing a gameto reduce lonelinessAcademic: McCann, LisaProject is suitable for 2 studentsThere have been increased numbers of people who have experienced mental health challenges dueto the significant periods of isolation as a consequence of the global Covid-19 pandemic andassociated lockdown measures. Digital Health could provide solutions to support children duringperiods of isolation or separation from peers due to illness to help them during periods of loneliness.This project would focus on developing a fun game for primary school aged children to play thathas an underlying health focus but delivered in a fun way via a digital game that they can play withtheir peers and friends (drawing on the serious games concept). Ideally the game would havemechanisms that allow children to record and report their feelings and emotions via game play,connect safely to peers and for a companion flagging system/app for parents for any reports ofconcern the children report via the game play. This project would suit students with advancedsoftware and or game development skills or students with a keen interest in user design and userexperience and an interest in using games for health benefits and engaging a younger audience inreporting on their own health status.MCCA09 – Supporting young people with cancer and their parents aftertreatmentAcademic: McCann, LisaYoung people diagnosed with cancer face a number of unique challenges particularly thoseassociated to their developmental and life-course transitions including body image, education,employment, independence, sexuality, and impact on relationships. The end of active treatment canactually be a particularly anxious time for young people and their parents as their contact withmedical professionals decreases but the physical and psychological impact of the cancer diagnosisand its treatment remain ever present or worsen. The end of treatment is not only stressful andanxiety inducing for young people but their parents too as their involvement in the care pathwayevolves throughout the young person’s illness experience. The aim of this project is to rapidlydevelop prototype mobile applications to provide supportive care for young people with cancer andtheir parents in the transition to treatment completion, utilising the concept of remote monitoring.There are three parts to this project – an app for young people to use, a companion app for theparents to use and a dashboard of information presented to and accessible by the health professionalteam. This project would be suitable for students with strong desk-based research skills and studentswith an interest in developing prototype apps at medium-high fidelity level based on coding skillsand knowledge.MCCA10 – Development of an exercise-based app to support recovery frommajor abdominal surgeryAcademic: McCann, LisaMajor abdominal surgery can take a person many weeks and months to recover from, but dailygradual movements are important to help aid the recovery process. However, people for whom largeincisions in the abdomen are required, compared to those who receive key hole surgery, can have alonger and more complicated recovery periods. Thus they may have requirements for appropriatelysupported and advised exercise programmes. This project would focus on co-designing and creatinga prototype app to support this population with easy to follow guided exercises, tips and strategiesfor rehabilitation and diaries or trackers to monitor progress. This project would suit a studentinterested in fitness and exercise, surgical interventions, co-designing and developing health apps atprototyping / coded prototyping levels.MCCA11 – Becoming an Ostomate – adjusting to life with a stomaAcademic: McCann, LisaProject is suitable for 2 studentsPeople who undergo major and life-altering surgery to have a stoma created can experience a rangeof lifestyle and life adjustments as they adapt to their post-surgery life with their colostomy orileostomy. Such changes can include changes to diet, exercise, sleep, mental health and wellbeing,confidence, self-esteem and vastly altered self-care routines and requirements. This project wouldfocus on developing a supportive intervention to help people, particularly older people, who haverecently become an Ostomate with these mental and physical changes so the intervention needs tobe accessible, simple to use and accessible on different platforms, i.e. web app and smart phoneapp. This project would suit students keen to focus their work on older people, address a realclinical need and develop a person-centred prototype intervention.MCCA13 – Supporting the healthcare needs of remote and rural populations viadigital interventionsAcademic: McCann, LisaFor populations who live in remote and rural locations, there are challenges with accessinghealthcare services equitably compared to those who live closer to hospital and care services. Thereare a range of digital solutions and interventions that can help support these individuals includingremote monitoring technologies. In this project, the student will be required to shape the focus forthis work based on their own interests in terms of populations, contexts, technologies and health andcare issues and will be required to design and develop a prototype application/intervention/softwarethat would help address this issue. Therefore, this project would be best suited to a student who hasan interest in remote and rural healthcare issues and has an interest in developing a high fidelityprototype (coded or not) to address a relevant need for such an intervention – which they should beinformed by from their literature review.MCCA14 – Development of an accessible solution to support health outcomes ofpeople released from prisonAcademic: McCann, LisaPeople in prison experience high rates of complex mental health problems, substance use disordersand previous life adversities. It is often the most vulnerable and disadvantaged groups in societywho face digital exclusion, including older people, people with disabilities, those with lower wages,prisoners and former prisoners (UK Government, 2014). Upon release from prison, people continueto experience poor health and social outcomes and are also impacted by the digital divide – so thisproject will focus on the development of an accessible solution to support people with experience ofprison environments to record, monitor and share their health and wellbeing status with relevantprofessional support staff. This project involve the student using a range of recognised co-designtools and methods to create and evaluate the product. This project would suit a student interested inthe needs of vulnerable populations and utilising their own skillsets in the HCI domain to developthe prototype for an accessible and meaningful platform to capture the health and care needs ofpeople.MCCA15 – Becoming an Ostomate Carer – adjusting to life caring for someonewith a stomaAcademic: McCann, LisaPeople who undergo major and life-altering surgery to have a stoma created can experience a rangeof lifestyle and life adjustments as they adapt to their post-surgery life with their colostomy orileostomy. Such changes often directly impact on the lives of loved ones of the Ostomate as theyadopt caring roles to support their loved one. This new role requires the carer to rapidly learn aboutstoma maintenance and care including cleaning, bag changing, health of the stoma and dietaryrequirements of their loved on. In addition, there can be impacts on the carers own sleep, mentalhealth and wellbeing, confidence, and independence given the vastly altered self-care routines andrequirements. This project would focus on developing a supportive intervention to help peoplecarers, particularly older carers, who have recently become an Ostomate Carer with their new caringroutine and requirements. The intervention needs to be accessible, simple to use and accessible ondifferent platforms, i.e. web app and smart phone app. This project would suit students keen tofocus their work on the needs of older people and develop a person-centred prototype intervention.MCCA16 – Design and development of a digital health intervention for usewithin the context of evolving digital health landscapes and infrastructuresAcademic: McCann, LisaProject is suitable for 2 studentsThere are variations in the digital health landscape and readiness for digital health interventionsacross the world and there are influencing human, technology and systems factors relating to thedesign and development of these to make them accessible, ready and adoptable by stakeholders andsystems. This project will focus on designing and developing a digital health intervention targetedspecifically within this context – but this context, type of intervention, population and purpose canbe driven by the student in consultation with the supervisor. This project would be suitable forstudents with an interest in health interventions, contexts beyond the UK and for those students whoare required to develop and evaluate a prototype application for their dissertation requirements.NAIM06 – Abnormal Activity Detection using Deep Learning at Nursing HomeAcademic: Muhamad Naim, NurA nursing home is a facility for the residential care of elderly or disabled people. Residential carehomes help people manage daily life, such as assisting with getting dressed, washing, and eating.However, the trained care staff is not always by their sides. Thus real-time monitoring is veryimportant to improve the effectiveness.This project will focus on abnormal activities detection at the care home. Selected imagesdescribing fall and static movement (or any abnormal activity) will be trained using deep learning.The real-time monitoring system will identify normal or abnormal. The challenge is to propose anew deep learning algorithm that can understand and differentiate between negative and positiveactions at nursing homes. This project needs to understand the normal activities so that it can detectabnormal activities.Requirements: Deep Learning, PythonNAIM07 – Mental Health and Wellbeing Monitoring using Type-2 FuzzyLearning SystemAcademic: Muhamad Naim, NurMental illness is one of the major public health challenges in Scotland. Around one in three peopleare estimated to be affected by mental illness in any one year (https://www.gov.scot › policies ›mental-health, 15/08/2021). A system will be developed for individual monitoring. This system iscreated for people to takeaction before it gets worse.Type-2 Fuzzy Logic System is an algorithm that utilizes rule-based techniques for decision-making.All criteria from the literature will be considered to construct fuzzy rules such as a happy intimaterelationship with a partner, the network of close friends, an enjoyable and fulfilling career, enoughmoney, regular exercise, nutritional diet,enough sleep, and etc. Surveys will be conducted to collect information to propose a new fuzzyrules algorithm.Requirements: Fuzzy Logic SystemNAIM08 – Learning Rule-based for Smart Home Alert System and EnergySavingAcademic: Muhamad Naim, NurThere are many security systems that have been developed utilising CCTV to monitor smart homes.However, the criminals have aware of CCTV existence and mostly the images are not able to helpfor tracing or even in real-time monitoring. This research is based on fuzzy learning rule-based todetect abnormality at smart homes.A smart home usually consists of embedded sensors for autonomous actions. Automatic electricalappliances such as air-conditioners, window blinds, bulbs, doors, and etc. This project will combineall smart appliances in a smart home using a simulator to construct learning rule-based by a user.User activities will be learned to detect abnormality and at the same time be able to save energy.Requirement: Fuzzy Logic SystemNAIM09 – Fuzzy – Deep Learning Investigation in COVID-19 Detection basedon Chest X-ray Images and Clinical DatasetAcademic: Muhamad Naim, NurCurrent research mostly focuses on COVID-19 detection using deep learning algorithms from chestx-ray images. In this research, we will integrate chest x-ray images and clinical datasets to improvethe accuracy of detection. The study aims to increase efficiency by adopting a Fuzzy InferenceSystem into deep learning architecture to solve the higher level of uncertainties. Following are linksshowing potential open-source datasets that might be used for the research.https://github.com/ieee8023/covid-chestxray-datasethttps://www.kaggle.com/einsteindata4u/covid19/version/7Requirement: Deep Learning, Fuzzy Logic System, PythonNAIM10 – A Hybrid Algorithm to Detect and Recognise FacesAcademic: Muhamad Naim, NurSuitable for: ACSYolo is a very powerful object detection algorithm. It can detect faces in far distances quiteefficiently and recognise that is human. However, it is not able to recognise any particular person.Thus, this study will focus on integrating two algorithms which are Yolo and face recognisationalgorithms.There are many popular face recognition algorithms such as MTCNN, Dlib, and CVlib, thus acomparison study will be made to choose the best algorithm to integrate with Yolo. Differentvariables will be investigated such as distance and lighting.Requirements: Python and Deep learningNAIM11 – Text Documents Clustering Using Deep Learning AlgorithmsAcademic: Muhamad Naim, NurSuitable for: ACSText document clustering (TDC) represents a key task in text mining and unsupervised machinelearning, which partitions a specific documents’ collection into varied K-groups according to certainsimilarity/dissimilarity criteria. There exists a considerable amount of knowledge in the textclustering field and many attempts were carried out to resolve the TDC problem and improve thelearning performance.This study will focus on implementing deep learning algorithms for text documents clustering.Popular algorithms will be tested such as RNN and LSTM. A hybrid algorithm will be proposed toincrease the performance.Requirements: Deep Learning and Python.Reference:https://www.sciencedirect.com/science/article/pii/S1568494619307847NAIM12 – Power Scheduling in Smart Homes Using Deep Learning AlgorithmsAcademic: Muhamad Naim, NurSuitable for: ACSOptimizing the power demand for smart home appliances in a smart grid is the primary challengefaced by power supplier companies, particularly during peak periods, due to its considerable effecton the stability of a power system. Therefore, power supplier companies have introduced dynamicpricing schemes that provide different prices for a time horizon in which electricity prices arehigher during peak periods due to the high power demand and lower during off-peak periods. Theproblem of scheduling smart home appliances at appropriate periods in a predefined time horizon inaccordance with a dynamic pricing scheme is called the power scheduling problem in a smart home(PSPSH).Thus this study will focus on developing PSPSH using deep learning algorithms. Few deepalgorithms will be tested to find the highest accuracy. Hybrid algorithms will be proposed toincrease the accuracy.Requirements: Python and Deep LearningReference:https://www.sciencedirect.com/science/article/pii/S1364032119305702NAIM13 – Integration of Voice and Face Deep Learning Detection based onFuzzy Logic System for Identification SystemAcademic: Muhamad Naim, NurSuitable for: ACSRecently, identification system has been widely used at entrances such as at smart homes andfactories. Mostly, the system are using only face detection where usually needs clear images andlighting to identify users. The system is usually set up outdoor and has a high chance of havingerrors detection. Thus this study tries to integrate voice and face detection for identification.Deep learning algorithms will be used for voice and face detections. Fuzzy logic rule-based willdetermine the identification. The system will be tested with different deep learning and fuzzyalgorithms to increase accuracy.Requirements: Python, Deep Learning and Fuzzy ModulesNAIM14 – Fuzzy Logic System with Rule-Based Categorization in UncertainEnvironementsAcademic: Muhamad Naim, NurUncertainty evaluation among humans is completely not presented in shared intelligentenvironments. Human judgment and perception are slanted and intuitive. Although IntelligenceComputing is intended to mimic human actions, however, it still struggles to compute humanuncertainties.As a result, the Fuzzy Logic System (FLS) has been rapidly developed over the years to overcomethis problem using rule-based reasoning. FLS is also known as an intelligent algorithm that is ableto value abstract information such as feeling and emotion. Rules-based are generated based oncategorisation to reason human behavior and understanding, which are the core elements for thisresearch. Open source datasets will be selected to test the algorithms.Requirement: FLSNAIM15 – A New General Type-2 Fuzzy Logic SystemAcademic: Muhamad Naim, NurThe fuzzy logic system has tremendously developed over the years. In 1998, Karnik and Mendelproposed Type-2 Fuzzy Logic System (T2FLS). Recently, several researchers have started toexplore the application of General Type-2 Fuzzy Sets (GT2FS) and systems. GT2FS, which is oneof the advanced extensions of type-2 fuzzy sets to capture higher levels of uncertainties. However,type-2 fuzzy sets are more difficult to use and understand than are type-1 fuzzy sets (conventional);hence, their use is not yet widespread and computationally expensive.To simplify the GT2FS, firstly we intend to model preferences/perceptions from humans by usingGT2FS. Currently, we have a membership function for specific terms/words involving numericalevaluation such as temperature, age, humidity, happiness, etc. However, the proposed studysuggests developing a new way to compute human preferences by using the third dimension ofgeneral type-2 fuzzy sets. Secondly, to develop GT2FLS with human preferences as additionalparameters. Finally, to test the new algorithm with open-source datasets.Requirement: Fuzzy Logic SystemOTHER01 – Generating abstractive summaries for news paper articlesAcademic: Pathak, KanaadFull details of the the proposal can be found here: https://coordinated-drug-ce2.notion.site/MastersProject-Proposal-800398d7157a4193be0efcce7486ea19OTHER02 – Kappa language front-end to Algebraic Julia and CatlabAcademic: Waites, WilliamThe Kappa calculus for rule-based modelling is a stochastic process calculusoriginally intended for studying molecular biology. The canonicalimplementation is written in Ocaml [1]. It turns out to also be very wellsuited to the study of epidemics [2,3] enabling succinct expression of richdynamics. The Algebraic Julia project aims to build up practical scientificcomputing infrastructure from category theoretic concepts and intra aliasupports the formulation of models as Petri nets which can be viewed as aspecial case of the rule-based graph rewriting of Kappa.This project is to enable the expression of rule-based models in the Kappalanguage for execution in the Algebraic Julia system. It will involve creatinga parser for the language and a code generator for the correspondingrepresentation in Julia. Some experience with the Julia language is requiredand some familiarity with category theory is advantageous (an excellentintroduction from a programmer’s perspective is on this blog [6] and theaccompanying lecture series).[1] The Kappa platform for rule-based modeling https://doi.org/10.1093/bioinformatics/bty272[2] Rule-based epidemic modelshttps://www.sciencedirect.com/science/article/pii/S0022519321002708[3] Compositional modelling of immune response and virus transmission dynamicshttps://arxiv.org/abs/2111.02510[4] Algebraic Julia https://www.algebraicjulia.org[5] Open Petri Nets https://arxiv.org/abs/1808.05415[6] Category Theory for Programmershttps://bartoszmilewski.com/2014/10/28/category-theory-for-programmers-the-preface/OTHER03 – Dynamic network support in the Extended Kappa languageAcademic: Waites, WilliamThe Kappa calculus for rule-based modelling is a stochastic process calculusoriginally intended for studying molecular biology. The canonicalimplementation is written in Ocaml [1]. It turns out to also be very wellsuited to the study of epidemics [2,3] enabling succinct expression of richdynamics. It is not, however, without its limits. In particular, in order tostudy epidemics on networks [4], it was necessary to create some extensions tothe language. These are implemented in the ExtKappa module of the NetABCpackage [5].The ExtKappa implementation is incomplete, providing just enough functionalityto support the use case of the above mentioned paper [4]. In particular,addition and deletion of edges is not implemented though it is accepted by thelanguage. Accomplishing these operations efficiently is challenging. Applyinga graph rewriting rule means determining all ways the rule can match. Thisentails enumerating subgraph isomorphisms which is an expensive operation. Agood strategy is to do the expensive operation once and thereafter compute theincremental changes induced by rewriting rules. This project is to do that andthereby enable the study of processes on dynamic networks. Experienceprogramming in Python is required.[1] The Kappa platform for rule-based modeling https://doi.org/10.1093/bioinformatics/bty272[2] Rule-based epidemic modelshttps://www.sciencedirect.com/science/article/pii/S0022519321002708[3] Compositional modelling of immune response and virus transmission dynamicshttps://arxiv.org/abs/2111.02510[4] Transmission dynamics of SARS-CoV-2 in a strictly-Orthodox Jewishcommunity in the UK https://www.medrxiv.org/content/10.1101/2021.10.28.21265615v1[5] NetABC https://git.sr.ht/~wwaites/netabcOTHER05 – Serial composition of computational models in NetABCAcademic: Waites, WilliamThe NetABC package [1] is designed to provide a high-level interface to commonoperations on computational models. It is motivated by infectious diseasemodelling but is by no means limited to that field. It supports fitting modelsto time-series or point data using Approximate Bayesian Computation (hence thename), sampling trajectories from stochastic models and saving data in acommon format (HDF5), and some postprocessing of data and simple plotting.This is accomplished in a way that is agnostic about the specific formulationof the models themselves: they may be posed as differential equations, writtenin the Kappa language, expressed as processes on networks, or indeed writtenin any way so long as a simple calling convention is maintained.An interesting thing that we might like to do is to combine models when theycapture processes that occur on different time scales. An example is acycle of yearly influenza epidemics and vaccination [2,3] whichcontains two models: each year, decisions about vaccination are made and anepidemic develops, decisions in one year depend on the outcome of the epidemicin the previous year. We might, however, wish to use a different epidemicmodel or a different decision model. A much more complex example is Karr’sWhole Cell Model [4] which is a composition of several dozen such submodels.This project is to implement and demonstrate a generic mechanism forconstructing such serial model compositions in NetABC.[1] NetABC https://git.sr.ht/~wwaites/netabc[2] Modeling the interplay between seasonal flu outcomes and individualvaccination decisions https://arxiv.org/abs/2101.07926[3] The influence of altruism on influenza vaccination decisionshttps://royalsocietypublishing.org/doi/10.1098/rsif.2012.0115[4] A whole-cell computational model predicts phenotype from genotypehttps://dx.doi.org/10.1016/j.cell.2012.05.044OTHER06 – CompuCell3D and MechanicaAcademic: Waites, WilliamProject is suitable for 5 studentsThese projects are together with colleagues at the Biocomplexity Institute atIndiana University and consist of various computing and software developmentworks supporting the open-source CompuCell3D (https://compucell3d.org/) andMechanica (https://github.com/tjsego/mechanica/tree/develop) projects.CompuCell3D is a lattice-based modeling and simulation environment thatemploys the Cellular Potts model for agent-based, spatial modeling ofmulticellular biological systems. Mechanica is a lattice-free modeling andsimulation environment that employs classical molecular dynamics for modelingbiological, chemical and biophysics problems. Both CompuCell3D and Mechanicaare written in C++ and distribute Python language bindings.The projects below are each suitable for one student with strong C++ andpython programming ability.1. Mechanica multi-window application. Mechanica supports interactive,real-time simulation and provides real-time visualization of a model systemwhile a simulation is executed, which allows the user to interact withtheir model system and observe results as it evolves. However, Mechanicacurrently only supports exactly one rendering window, which can makeobserving relevant details of a simulation difficult. This project willdeliver the capability to provide multiple views of a simulation as it isexecuted, with programmatic and interactive interfaces for specifyingviewing details (e.g., number of windows, open window on event, etc.). Workwill include backend development of infrastructure for runtime managementof rendering contexts and window management, as well as frontenddevelopment of user control interfaces.2. Multi-GPU Mechanica deployment. Mechanica current implements a modulararchitecture of backend computational components that can be offloaded to asingle CUDA-supporting GPU, where users can select which computationalcomponents to offload based on the details of their simulation. Thisproject will further Mechanica’s modular architecture by enablingdeployment of backend components onto an arbitrary number of availableGPUs. Work will consist of backend development of supporting infrastructureto dynamically manage engine modules running across an arbitrary number ofGPUs, as well as some light development of parallel algorithms. Work willalso include some light frontend development of user interfaces formanaging GPU devices and engine components.3. Programmable Mechanica GUI and visualization components. Mechanica provides arobust and ever-expanding event system to deliver its promise of real-time,interactive simulations, including the user ability to write customkeyboard callbacks that trigger user-specified model events, or visualizesimulation data using third-party libraries (e.g., Mechanica + matplotilbin Python). This project will further Mechanica’s interactive userexperience by providing native GUI and data visualization components asbuilt-in, customizable features. Work will include backend integration anddeployment of relevant supporting libraries, and frontend development ofrelevant user-facing graphical interfaces to deploy customizable GUI andvisualization components.4. Mechanica integration in CompuCell3D. CompuCell3D can model a lot ofobjects and processes associated with subcellular, cellular andmulticellular processes. However, CompuCell3D currently is not particularlyuseful for modeling the effects of fluid and solid mechanics on cellularbehaviors. Using Mechanica’s ability to do particle-based modeling offluids and solids, this project will perform basic integration of Mechanicainto CompuCell3D as a basic, general simulation feature. Work will includebackend integration of Mechanica into the CompuCell3D core and lightdeployment supporting future development of solid and fluid mechanicssolvers in CompuCell3D.5. CompuCell3D lattice surface modeling. Diffusion and reaction of solublechemical species are critically important to biological organization, andCompuCell3D responds in kind by providing a suite of reaction-diffusionsolvers that can be configured and deployed in multicellular simulations.Being a lattice-based approach, CompuCell3D readily has the potential toalso include explicit cell surface dynamics. This project will develop thispotential by developing and implementing a coherent notion of latticesurfaces in CompuCell3D simulations. Work will include algorithmic andbackend development constructing new fundamental CompuCell3D structures, aswell as some development supporting future work developing surfacediffusion and active transport modeling and simulation capabilities.OTHER08 – Visualising causality in rule-based modelsAcademic: Waites, WilliamThe Kappa calculus for rule-based modelling is a stochastic process calculusoriginally intended for studying molecular biology. The canonicalimplementation is written in Ocaml [1]. It turns out to also be very wellsuited to the study of epidemics [2,3] enabling succinct expression of richdynamics. Rules have an inherent causal property: in order to apply, toreplace the graph pattern on the left-hand side of a rule with that on theright, the left-hand side must match in the state of the system. There existsa partial order on rules: the result of applying one rule may be required by asecond rule. This project is to create a tool visualise this partial order fora given rule-set so that we may better understand the causal relations betweenthe processes that make up a rule-based model. Rule-sets can be constructed asneeded as well as taken from the literature on molecular biology andepidemics. Experience with Haskell, Ocaml or Python is useful (there areparsers for Kappa in these languages) though not essential if the studentwishes to create a parser in their preferred language.[1] The Kappa platform for rule-based modeling https://doi.org/10.1093/bioinformatics/bty272[2] Rule-based epidemic modelshttps://www.sciencedirect.com/science/article/pii/S0022519321002708[3] Compositional modelling of immune response and virus transmission dynamicshttps://arxiv.org/abs/2111.02510OTHER09 – Video Machine Learning for Emotion RecognitionAcademic: Abel, AndrewEmotion recognition from facial expressions is an established field of research, and in recent years,this has been expanded on by using video data and machine learning, in particular a combination ofconvolution neural networks (CNNs) and recurrent neural networks (RNNs), can be used toestimate emotions and map them in the emotional space [1]. There are a number of features that canbe used for emotion recognition, including pose, psychological measures, and facial features. In thisproject, we propose to investigate state of the art deep learning models, such as ResNet or the VGGmodel [2] for video emotion recognition.This is primarily a research based machine learning project, focusing on the fundamentals, and isstrongly oriented towards the use of a CNN end-to-end method. The student will investigate theliterature, identify state of the art research, and use a deep learning model to generate results, whichwill be fully analysed.[1] Abdulsalam, Wisal Hashim, Rafah Shihab Alhamdani, and Mohammed Najm Abdullah. “Facialemotion recognition from videos using deep convolutional neural networks.” International Journalof Machine Learning and Computing 9.1 (2019): 14-19.[2] Lu, C., Zheng, W., Li, C., Tang, C., Liu, S., Yan, S., & Zong, Y. (2018, October). Multiplespatio-temporal feature learning for video-based emotion recognition in the wild. In Proceedings ofthe 20th ACM International Conference on Multimodal Interaction (pp. 646-652).OTHER10 – Software Annotation for Multimedia DatasetsAcademic: Abel, AndrewWhen we perform machine learning, labelled data is often a requirement, whether for ground truthcomparisons or labelling. For example, if we want to recognise speech, then we need to be able tolabel where in a sentence a word begins and ends. Or similarly, when we are analysing a video, wemay need to label a range of different things, such as if someone is looking up or down, if they areshaking their head, blinking, or if they are speaking or not. Depending on what we wish to analyse,there is an almost limitless range of potential labelling that may need to be done. This is a majorreason why people use specific datasets in machine learning. Annotation can be time consuming,many video and audio databases are not well labelled, and recording custom corpora becomeschallenging.The aim of this project is to create an application, primarily using Python with a web basedinterface (such as streamlit) that will take a video/audio file, visualise it, give the user controls overplaying/pausing etc. and will allow the user to customise the number and type of potential labels.While a video is playing, the user can add labels using a simple, customisable interface, and thensave the results as a file. This is primarily an application based project, and the focus will primarilybe on backend operation.OTHER11 – Parallel Video Signal ProcessingAcademic: Abel, AndrewFeature extraction can be used to extract detailed audio and visual information, and Gabor basedmethods have been developed which have been used for speech recognition [1]. These are a form ofedge detector and allow us to extract detailed mouth information, such as height, width and area ofa speaker’s mouth [2]. Using these techniques, we have successfully extracted features andproduced good lipreading results. However, these approaches work under the assumption that thereis a single speaker within the camera shot at all times. In the real world, people move, multiplepeople may appear in a camera shot, and there may also be unseen speakers who do not match thespeaker.In this project, we propose to develop a parallel processing video feature extraction technique.Using our Python based face identification and feature extraction technique, we wish to expand it towork for multiple speakers, and use parallel processing so that all features can be extracted andvisualised for multiple speakers simultaneously.This is a programming based project with some research applications, and will require knowledgeof programming. The key aspect is researching and developing parallel processing.OTHER12 – Multistream Audiovisual Speech RecognitionAcademic: Abel, AndrewFeature extraction can be used to extract detailed audio and visual information, and Gabor basedmethods have been developed which have been used for speech recognition [1]. These are a form ofedge detector and allow us to extract detailed mouth information, such as height, width and area ofa speaker’s mouth [2]. Using these techniques, we have successfully extracted features andproduced good lipreading results. We can combine these Gabor features with CNN based features,such as 3D-CNN features, and also the audio stream, to produce improved speech recognitionresults. There are several ways to fuse these different streams [3] [4], and we will investigate thesein this project, and aim to improve on the state of the art.In this project, the student will use a suitable database to extract all features, and then carry out anumber of machine learning experiments using Python. This is primarily a research based machinelearning project, focusing on the fundamentals. The research can also be conducted using Chineselanguage or English language data. Chinese visual speech recognition presents some distinctchallenges.[1] Zhang, Xuejie, et al. “Visual Speech Recognition with Lightweight Psychologically MotivatedGabor Features.” Entropy 22.12 (2020): 1367.[2] Xu, Yan, Yuexuan Li, and Andrew Abel. “Gabor based lipreading with a new audiovisualmandarin corpus.” International Conference on Brain Inspired Cognitive Systems. Springer, Cham,2019.[3] Petridis, Stavros, et al. “End-to-end audiovisual fusion with LSTMs.” arXiv preprintarXiv:1709.04343 (2017).[4] Katsaggelos, Aggelos K., Sara Bahaadini, and Rafael Molina. “Audiovisual fusion: Challengesand new approaches.” Proceedings of the IEEE 103.9 (2015): 1635-1653.OTHER13 – Robust Multistream Audiovisual Speech RecognitionAcademic: Abel, AndrewFeature extraction can be used to extract detailed audio and visual information, and Gabor basedmethods have been developed which have been used for speech recognition [1]. These are a form ofedge detector and allow us to extract detailed mouth information, such as height, width and area ofa speaker’s mouth [2]. Using these techniques, we have successfully extracted features andproduced good lipreading results. We can combine these Gabor features with CNN based features,such as 3D-CNN features, and also the audio stream, to produce improved speech recognitionresults. There are several ways to fuse these different streams [3] [4], and previous project work hasproduced results. However, these systems assume that all streams are available at all times. Theremay be cases in the real world where not all streams are available, for example, in communication,if audio noise is present, or if the speaker turns away and visual information can not be extracted.The aim of this project is to experiment with different scenarios and identify how performance isaffected by lack of availability.In this project, the student will use a suitable database to extract all features, and then carry out anumber of machine learning experiments using Python. This is primarily a research based machinelearning project, focusing on the fundamentals. The research can also be conducted using Chineselanguage or English language data. Chinese visual speech recognition presents some distinctchallenges, for several reasons. Firstly, much research has been focused on English languageresearch, and Chinese language is a tonal language formed of distinct characters.[1] Zhang, Xuejie, et al. “Visual Speech Recognition with Lightweight Psychologically MotivatedGabor Features.” Entropy 22.12 (2020): 1367.[2] Xu, Yan, Yuexuan Li, and Andrew Abel. “Gabor based lipreading with a new audiovisualmandarin corpus.” International Conference on Brain Inspired Cognitive Systems. Springer, Cham,2019.[3] Petridis, Stavros, et al. “End-to-end audiovisual fusion with LSTMs.” arXiv preprintarXiv:1709.04343 (2017).[4] Katsaggelos, Aggelos K., Sara Bahaadini, and Rafael Molina. “Audiovisual fusion: Challengesand new approaches.” Proceedings of the IEEE 103.9 (2015): 1635-1653.OTHER14 – NetABC in an supercomputing environmentAcademic: Waites, WilliamThe NetABC package [1] is designed to provide a high-level interface to commonoperations on computational models. It is motivated by infectious diseasemodelling but is by no means limited to that field. It supports fitting modelsto time-series or point data using Approximate Bayesian Computation (hence thename), sampling trajectories from stochastic models and saving data in acommon format (HDF5), and some postprocessing of data and simple plotting.This is accomplished in a way that is agnostic about the specific formulationof the models themselves: they may be posed as differential equations, writtenin the Kappa language, expressed as processes on networks, or indeed writtenin any way so long as a simple calling convention is maintained.Many of the tasks that NetABC accomplishes are embarassingly parallel andwell-suited to multi-processor environments. Fitting models to data involvesrunning a model many times with different parameter values. Samplingtrajectories from a stochastic model simply involves running it many times. Asimplemented, NetABC can only do this on a single host. This project is toextend the implementation to work in a distributed memory high performancecomputing setting. Such supercomputers are batch processing systems workingwith a job scheduler (typically Slurm). Access will be provided to sucha system for this project.[1] NetABC https://git.sr.ht/~wwaites/netabcOTHER15 – Deep Learning/Machine Learning models for Video Coding forMachinesAcademic: Fernando, AnilProject is suitable for 6 studentsDue to the large storage and bandwidth needed, video must be compressed before they are storedand/or transmitted. Video compression has been fully optimised for human consumption. Unlikehuman vision, machine vision has different purposes and evaluation metrics in consuming video.Therefore, current video codecs are not suitable for compressing the video for machines. To solvethis interesting and challenging problem, Video coding for machine (VCM) will be required. Thisapproach assists to correlate feature coding for machine vision and video coding for human vision.VCM will be a key solution to the most severe challenges of multimedia computing, transmission,and storage, as well as to the rise of AI-driven video intelligent solutions. As a result of VCM,everyday video content will be transformed by identifying, classifying, and indexing objects thatappear in it, making the metadata machine specific, searchable, and actionable. In this project, DeepLearning/Machine Learning models are explored in compressing the video for machines. Differenttarget applications will be considered.OTHER16 – Explainable AI Methods for Identifying the Critical Areas of ChestX-ray ImagesAcademic: Fernando, AnilProject is suitable for 2 studentsConvolutional neural networks (CNNs) are used in identifying diseases in chest using chest x-rayimages. However, experts in the medical field reluctant to accept the decisions made by CNNs asthey are black-box models and cannot explain how the model arrived at the conclusion. Recently,different explainable AI methods such as LIME, Grad-CAM, Guided Grad-CAM IntegratedGradients, Guided Integrated Gradients have been introduced to produce visual explanations fordecisions made from the CNN models. In this research project, a comparative study will be carriedout to compare these methods on chest x-ray images when identifying the chess related diseases andthe critical areas in these chest x-ray images using convolutional neural networks.OTHER17 – Explainable AI for Retinopathy predictionAcademic: Fernando, AnilDeep learning models are used to predict Retinopathy from retinal images. Though they providegood results, deep learning models are used as black-box models only. This creates trust issues onAI models and their decisions to the clinician and patients. To overcome this, Explainable-AI can beused as a model which can assist clinicians in evaluating the recommendations provided by the AImodel based on their experience and clinical judgment.OTHER18 – AI models for Retinopathy prediction using Non-Mydriatic fundusimagesAcademic: Fernando, AnilAI-based models are applied to predict Retinopathy from the mydriatic and light illuminated fundusimages. But, applying dilating drops needs to be done by a clinician to avoid complications, limitingthe home usage of retinopathy prediction tools. Nonmydriatic cameras make fundus photographyfar more patient-friendly by eliminating the need for bright lights and dilating drops. This studyfocuses on image processing and GAN based technique to predict Retinopathy from Non-Mydriaticfundus images that are rarely available in the public databases.OTHER19 – Attention Based Glaucoma DetectionAcademic: Fernando, AnilDeep learning models are used to predict Glaucoma using fundus images. As there is highredundancy in fundus images, an attention-based model can be applied to boost the performance ofGlaucoma prediction. This study applies the attention-based model and compares it with otherapproaches.OTHER20 – Subtitle Generator with LipreadingAcademic: Abel, AndrewProject is suitable for 2 studentsFeature extraction can be used to extract detailed audio and visual information, and Gabor basedmethods have been developed which have been used for speech recognition [1]. These are a form ofedge detector and allow us to extract detailed mouth information, such as height, width and area ofa speaker’s mouth [2]. Using these techniques, we have successfully extracted features andproduced good lipreading results. However, these are lab based techniques, and have not yet beentested in the real world. We propose to experiment with visual speech recognition to attempt tosubtitle videos. We want to test our algorithm on real videos (lecture data), and compare audio onlywith visual only results.In this project, the student will use a database of teaching videos to extract data, and then use thisdata to carry out a number of machine learning experiments using Python. This project is suitablefor two students. One student to focus more on generating a practical output (a subtitle file), andone student to work on the more research focused experiments.[1] Zhang, Xuejie, et al. “Visual Speech Recognition with Lightweight Psychologically MotivatedGabor Features.” Entropy 22.12 (2020): 1367.[2] Xu, Yan, Yuexuan Li, and Andrew Abel. “Gabor based lipreading with a new audiovisualmandarin corpus.” International Conference on Brain Inspired Cognitive Systems. Springer, Cham,2019.OTHER21 – Live Lipreading with Gabor Based Visual FeaturesAcademic: Abel, AndrewProject is suitable for 2 studentsFeature extraction can be used to extract detailed audio and visual information, and Gabor basedmethods have been developed which have been used for speech recognition [1]. These are a form ofedge detector and allow us to extract detailed mouth information, such as height, width and area ofa speaker’s mouth [2]. Using these techniques, we have successfully extracted features andproduced good lipreading results. However, these are lab-based techniques, and have not yet beentested in the real world.We propose to develop our first live lipreading system. The aim is that the system will receive avideo signal, divide it into words, and then process it in real time to generate real time results, sothat subtitling is performed on live videos. This project is suitable for two students. One to focus onthe live input and data processing (the software development part), and one to be more focused onthe research aspect (lip reading using machine learning, comparisons with audio speechrecognition).[1] Zhang, Xuejie, et al. “Visual Speech Recognition with Lightweight Psychologically MotivatedGabor Features.” Entropy 22.12 (2020): 1367.[2] Xu, Yan, Yuexuan Li, and Andrew Abel. “Gabor based lipreading with a new audiovisualmandarin corpus.” International Conference on Brain Inspired Cognitive Systems. Springer, Cham,2019.OTHER22 – Audiovisual Speech SynthesisAcademic: Abel, AndrewFeature extraction can be used to extract detailed audio and visual information, and Gabor basedmethods have been developed which have been used for speech recognition [1]. These are a form ofedge detector and allow us to extract detailed mouth information, such as height, width and area ofa speaker’s mouth [2]. As well as speech recognition, these methods can also be used for speechsynthesis where we use visual information to generate an estimated audio signal, which is importantfor potential noise removal. We have some very preliminary results, and there is considerable roomfor improvement. In this project, we will recreate the initial work and then try to improve thesynthesis system. The student will work with speech autoencoding, feature extraction, machinelearning, and the student will work with the existing code, and will develop it further for improvedusability.[1] Zhang, Xuejie, et al. “Visual Speech Recognition with Lightweight Psychologically MotivatedGabor Features.” Entropy 22.12 (2020): 1367.[2] Xu, Yan, Yuexuan Li, and Andrew Abel. “Gabor based lipreading with a new audiovisualmandarin corpus.” International Conference on Brain Inspired Cognitive Systems. Springer, Cham,2019.OTHER23 – Adaptive Eye Blinking MeasurementAcademic: Abel, AndrewFacial expressions associated with the upper face, in particular the eyes, are extremely useful inemotion recognition, with numerous potential applications such as medical diagnostics and humancomputer interaction. There are a number of useful features, including eye blinking, winking, andeyebrow movement. We are particularly interested in eye blinking, which can be divided intoinvoluntary blinking and voluntary blinking. Voluntary blinking is associated with means ofcommunication, such as interfaces to computing devices. Detected blinking can serve as anadditional interface with appropriate responses for a specific action. Involuntary blinking can beused as a criterion to diagnose medical conditions such as Tourettes syndrome, disorders of thenervous system can be associated with excessive blinking, and alternatively a reduced rate ofblinking is associated with Parkinson’s disease. There are a number of different approaches to detecteye blinking. We have proposed a new approach to detected blinking by Lacrimal aspect ratio. Inwhich the lacrimal area is used to detect blinking and eyebrow movement.We have found that subject distance from camera can affect this measurement, as the area changeswith movement. As a result, lacrimal area may affect the accuracy of the blinking. In order to makeit more adaptable, we propose to evaluate the effects of different eye movements and eye blinkingwith respect to the distance from camera to improve accuracy. We also want to compare their effectson different light intensities and different conditions along with various subjects including, oldpeople, male female and children and compare their result and effects on blinking while they arenear or far from camera. Finally, we plan to investigate the development of an adaptive approach toimprove accuracy. This is a python and image processing project, with some research aspects, youwould be working with existing code, and improving it.OTHER24 – Gabor Based Eye measurementsAcademic: Abel, AndrewFacial expressions associated with the upper face, in particular the eyes, are extremely useful inemotion recognition, with numerous potential applications such as medical diagnostics and humancomputer interaction. There are a number of useful features, including eye blinking, winking, andeyebrow movement. We are particularly interested in eye blinking, which can be divided intoinvoluntary blinking and voluntary blinking. Voluntary blinking is associated with means ofcommunication, such as interfaces to computing devices. Detected blinking can serve as anadditional interface with appropriate responses for a specific action. Involuntary blinking can beused as a criterion to diagnose medical conditions such as Tourettes syndrome, disorders of thenervous system can be associated with excessive blinking, and alternatively a reduced rate ofblinking is associated with Parkinson’s disease. There are a number of different approaches to detecteye blinking. We have proposed a new approach to detected blinking by Lacrimal aspect ratio. Inwhich the lacrimal area is used to detect blinking and eyebrow movement.Feature extraction can be used to extract detailed audio and visual information, and Gabor basedmethods have been developed which have been used for speech recognition [1]. These are a form ofedge detector and allow us to extract detailed mouth information, such as height, width and area ofa speaker’s mouth [2]. Using these techniques, we have successfully extracted features andproduced good lipreading results. It should be theoretically possible to use these features to extractother regions of the face, in particular, the eye regions. However, we have not yet tested this. Theaim of this project is to test and evaluate the use of Gabor features for generating measurements ofeye movements. We can then compare our results to other projects. This is a reasonably open endedproject, with some flexibility, and is based on Python programming with some image processing.[1] Zhang, Xuejie, et al. “Visual Speech Recognition with Lightweight Psychologically MotivatedGabor Features.” Entropy 22.12 (2020): 1367.[2] Xu, Yan, Yuexuan Li, and Andrew Abel. “Gabor based lipreading with a new audiovisualmandarin corpus.” International Conference on Brain Inspired Cognitive Systems. Springer, Cham,2019.ROPE10 – Self-Healing Programs and Self-Adaptive SystemsAcademic: Roper, MarcProject is suitable for 3 studentsA long-term vision within computer science is the idea of Autonomic Computing – the ability tobuild systems which essentially look after themselves and respond to any internal or externalchanges by being self-monitoring, self-adapting etc. (conveniently summarised to self-*). In a smallbut significant step towards this goal, a seminal paper by Weimer et al. describes a system that takesa program along with a number of passing and failing test cases and uses evolutionary computationto search for a repair to the program – i.e. automatically transforms the code so that all the test casesnow pass. This has now beccome a very active area of research. The aim of this project is toresearch some of the more recent developments in the area and explore building a similar selfhealing system in Java (or similar language). Constructing a full system is beyond the scope of thisproject so you could focus on certain aspects of a system, language or domain for instance.Thers is also the possibility to broaden the scope of the project by considering how evolutionarytechniques may be used to control and adapt the execution of a system in response to external andinternal pressures. For instance, load balancing is a typical example of this but very interestingdevelopments have been taking place in the robotics area with efforts to evolve robots that are ableto adapt to their environments.This project is able to support a number of students if they are able to take the ideas and principlesoutlined above and explore their application to different problems or use very different techniques.Reference:Weimer, W., Forrest, S., Le Goues, C., and Nguyen, T. 2010. Automatic program repair withevolutionary computation. Commun. ACM 53, 5 (May. 2010), 109-116. DOI=http://doi.acm.org/10.1145/1735223.1735249ROPE11 – Ontology-Based Evolutionary ComputationAcademic: Roper, MarcEvolutionary Computation or Evolutionary Algorithms are a class of algorithms inspired by theprinciples of natural selection and designed to solve a broad range of optimisation problems.Ontologies are codification of a concept which specifies the relationship between entities, objectsand concepts. There is a lot of interest in the way that ontologies and evolutionary computing maybe combined – for instance an ontology can provide useful information about a domain or even apotential mechanism for decomposing a fitness function. The aim of this project is to explore theuse of ontologies to support evolutionary computing. The suggested approach (similar to thatdescribed in the paper by Wimmer and Rada – see below for full reference) would involve selectingone or more candidate problems and developing a system which makes use of both concepts andexplores their interaction, but other alternatives could be investigated.Hayden Wimmer, Roy Rada, “Good versus bad knowledge: Ontology guided evolutionaryalgorithms”, Expert Systems with Applications, Volume 42, Issue 21,2015, Pages 8039-8051, ISSN0957-4174,https://doi.org/10.1016/j.eswa.2015.04.064.(https://www.sciencedirect.com/science/article/pii/S0957417415003061)ROUS01 – Interactive Vizualization for Transformer ModelsAcademic: Roussinov, DmitriProject is suitable for 3 studentsAttention-based Transformers are currently enjoying great interest from the researches and are usedin the applications that we use every day on the Web and mobile devices such as Audio input,machine translation, playing games, face identification, and e-mail spam filtering. This project willproduce an application (or set of useful programs) that would allow students to better understandhow Transformers operate, how we can train and use them, and what are their limitations. You willdecide on specific set of features to implement. One possibility will be involving Jupiter notebooks.More than average programming experience in Java or Python or C++ is highly desirable. This maybe experimentation-based or software development-based project. Will require plenty ofexperimentation. Do not expect to be given risk-free existing solutions. You will be expected to findthem yourself. At least 5 research papers on the topics need to be chosen, read and at leastsomewhat understood by the student.ROUS02 – Fictional Story GeneratorAcademic: Roussinov, DmitriProject is suitable for 3 studentsThe project will experiment with the existing methods and tools for automated generation ofcreative literature works (essays, stories, novels, blogs, etc.). One specific explored approach can beto use pre-trained unsupervised models such as T5 or GPT-2. There is no known solution to thisproblem, so it will be an experimentation based project. We may start with very simple 5-6 sentencestories, so the challenge will be to make those sentence connected into a single story line.Approaches based on attention-based transformers recently shown very effective for manyapplications is one particular possible approach but there are others.More than average programming experience in Java or Python or C++ is expected. You will need tolearn how to use PyTorch or Tensorflow library for deep learning. Not to be afraid of Linux shellcommand language. Will require plenty of experimentation, downloading and installing freeresearch software from the Web (github). You can re-use existing implementations so you don’thave to implement all from scratch, but you are expected to experiment with it and improve existingsolutions or create original code that manipulates the training data. Do not expect to be given riskfree existing solutions. You will be expected to find them yourself. At least 5 research papers on thetopics need to be chosen, read and at least understood by the student.ROUS03 – Generating Videos from ExamplesAcademic: Roussinov, DmitriProject is suitable for 2 studentsIn not too distant future, you will not always need actors and actresses to make a film. Thealgorithms are currently developed and tested to synthesize images and videos based on theexamples of existing works. The approach will involve training a deep neural network usingexamples of images and videos in a way similar to how the human imagination produces them. Thisis not to be confused with computer graphics, where the image is generated by geometric and lightmodelling.More than average programming experience in Java or Python or C++ is expected. You will need tolearn how to use Theano or Tensorflow library for deep learning. Not to be afraid of Linux shellcommand language. Will require plenty of experimentation, downloading and installing freeresearch software from the Web (github). You can re-use existing implementations so you don’thave to implement all from scratch, but you are expected to experiment with it and improve existingsolutions or create original code that manipulates the training data. Do not expect to be given riskfree existing solutions. You will be expected to find them yourself. At least 5 research papers on thetopics need to be chosen, read and at least somewhat understood by the student.ROUS04 – Catching Generated ContentAcademic: Roussinov, DmitriProject is suitable for 2 studentsNormal social network and microblog operation is often disrupted by “trolling” activities, includingthose sponsored by political opponents. This project will develop technologies to automaticallydetect the content that trolls currently generate (or can generate in future) using AI-tools such asGPT-2 and post in social media (e.g. Twitter, Facebook, Instagram). You can use deep neuralnetworks or suggest your own approaches.More than average programming experience in Java or Python or C++ is expected. You will need tolearn how to use Pytorch or Tensorflow library for deep learning. Not to be afraid of Linux shellcommand language. Will require plenty of experimentation, downloading and installing freeresearch software from the Web (github). You can re-use existing implementations so you don’thave to implement all from scratch, but you are expected to experiment with it and improve existingsolutions or create original code that manipulates the training data. Do not expect to be given riskfree existing solutions. You will be expected to find them yourself. At least 5 research papers on thetopics need to be chosen, read and at least somewhat understood by the student.ROUS05 – Clinical Events Predictions Based on TextAcademic: Roussinov, DmitriProject is suitable for 2 studentsThe project will explore how certain medical events (E.g. re-admitting a patient to a hospital, orrecovering from an illness) can be predicted based on the notes that the doctors write in free-textform. One specific explored approach can be to use pre-trained unsupervised models such as Bertand Elmo.More than average programming experience in Java or Python or C++ is expected. You will need tolearn how to use PyTorch or Tensorflow library for deep learning. Not to be afraid of Linux shellcommand language. Will require plenty of experimentation, downloading and installing freeresearch software from the Web (github). You can re-use existing implementations so you don’thave to implement all from scratch, but you are expected to experiment with it and improve existingsolutions or create original code that manipulates the training data. Do not expect to be given riskfree existing solutions. You will be expected to find them yourself. At least 5 research papers on thetopics need to be chosen, read and at least somewhat understood by the student.ROUS06 – Explanation capabilities of Pre-Trained Language ModelsAcademic: Roussinov, DmitriProject is suitable for 2 studentsPre-Trained Language Models, such as BERT, T5, GPT3 have been noted to have show capabilitiesto generate explanations if properly poked (e.g. by asking questions “why” in a simulated dialogue).However, those capabilities have not been well studied. You can start with using online demos, anddesign and carry-out interactive experiments to test those capabilities and possibly theirapplications. Further training (fine-tuning) of the existing models may be also involved if necessary.More than average programming experience in Java or Python or C++ is expected. You will need tolearn how to use PyTorch or Tensorflow library for deep learning. Not to be afraid of Linux shellcommand language. Will require plenty of experimentation, downloading and installing freeresearch software from the Web (github). You can re-use existing implementations so you don’thave to implement all from scratch, but you are expected to experiment with it and improve existingsolutions or create original code that manipulates the training data. Do not expect to be given riskfree existing solutions. You will be expected to find them yourself. At least 5 research papers on thetopics need to be chosen, read and at least somewhat understood by the student.ROUS07 – Teaching computers MathAcademic: Roussinov, DmitriProject is suitable for 2 studentsAlthough sounds very paradoxical since computers have been doing Math better than people fordozens of years, this problem recently came back, as people started to teach computers languagemodels, so they can understand and communicate with people in Natural Language. You can teach apre-trained language model (such as Bert, T5, GPT3 etc.) some Mathematical capabilities of yourchoice, even starting with such simple ones as counting or applying deduction. The most challengeis to be creative in where to find or how to artificially synthesize training data. Some recent researchexists on this topic. For DHS students, this can be in the E-health domain.More than average programming experience in Java or Python or C++ is expected. You may need tolearn how to use PyTorch or Tensorflow library for deep learning. Not to be afraid of Linux shellcommand language. Will require plenty of experimentation, downloading and installing freeresearch software from the Web (github). You can re-use existing implementations so you don’thave to implement all from scratch, but you are expected to experiment with it and improve existingsolutions or create original code that manipulates the training data. Do not expect to be given riskfree existing solutions. You will be expected to find them yourself. At least 5 research papers on thetopics need to be chosen, read and at least somewhat understood by the student.ROUS08 – Teaching computers how to ProgrammeAcademic: Roussinov, DmitriProject is suitable for 2 studentsExperiments with a GPT-3 (a language model) has shown that it can generate computerprogrammes based on descriptions written in natural languages, e.g. “a button that looks like awatermelon” or “large text in red that says WELCOME TO MY NEWSLETTER and a blue buttonthat says Subscribe.” You will explore how pre-trained language models such as GPT-2 or T5 cangenerate codes for simple applications of your choice. The most challenge is to be creative in whereto find or how to artificially synthesize training data. Some recent research exists on this topic.More than average programming experience in Java or Python or C++ is expected. You may need tolearn how to use PyTorch or Tensorflow library for deep learning. Not to be afraid of Linux shellcommand language. Will require plenty of experimentation, downloading and installing freeresearch software from the Web (github). You can re-use existing implementations so you don’thave to implement all from scratch, but you are expected to experiment with it and improve existingsolutions or create original code that manipulates the training data. Do not expect to be given riskfree existing solutions. You will be expected to find them yourself. At least 5 research papers on thetopics need to be chosen, read and at least somewhat understood by the student.ROUS09 – Teaching Transformers to Generate MusicAcademic: Roussinov, DmitriAttention-based Transformers are currently enjoying great interest from the researches and are usedin the applications that we use every day on the Web and mobile devices such as Audio input,machine translation, playing games, face identification, and e-mail spam filtering. The project willinvolve exploring how transformers can be used to generate music. You will need to implement asystem that can learn from examples, e.g. famous music pieces of a certain style, and will try toproduce something similar.More than average programming experience in Java or Python or C++ is expected. You may need tolearn how to use PyTorch or Tensorflow library for deep learning. Not to be afraid of Linux shellcommand language. Will require plenty of experimentation, downloading and installing freeresearch software from the Web (github). You can re-use existing implementations so you don’thave to implement all from scratch, but you are expected to experiment with it and improve existingsolutions or create original code that manipulates the training data. Do not expect to be given riskfree existing solutions. You will be expected to find them yourself. At least 5 research papers on thetopics need to be chosen, read and at least somewhat understood by the student.TERZ01 – An Online Version of OWASP CornucopiaAcademic: Terzis, SotiriosOWASP has developed Cornucopia (https://owasp.org/www-project-cornucopia/) a card game tohelp identify security requirements and promote developer awareness of security issues and securecoding techniques. The game has been designed to be played by developers sitting around a tablefocusing on a particular web application. However, this does fit well with development teams thatare not physically co-located. Supporting remote development teams is particularly important in acontext where working from home has become the norm. So, the aim of this project is to develop anonline version of the card game.Although the main focus is on developing the online version of the game, for students interested inmobile software security there is also the possibility of looking at developing a card deck coveringsecure mobile app development. Depending on the specific scope of the work, this is either asoftware-development based or an experimentation-based with significant software developmentproject.The project requires good software development skills and a strong interest in secure codingpractices.TERZ03 – ROSCA on EOS BlockchainAcademic: Terzis, SotiriosA rotating savings and credit association (ROSCA) is a group of individuals who agree to meet for adefined period in order to save and borrow together, a form of combined peer-to-peer banking andpeer-to-peer lending (see https://en.wikipedia.org/wiki/Rotating_savings_and_credit_association).EOS (https://eos.io/) is an open source blockchain-based platform intended to run decentralizedapplications and smart contracts.The aim of this project is to develop a ROSCA on EOS.A key challenge with this project will be to get familiar with EOS, its capabilities, as well asdeveloping applications in it.This is a software development based project. It requires strong programming skills in order toquickly get familiar with development on EOS.TERZ06 – Mobile Messaging PhishingAcademic: Terzis, SotiriosPhishing is a common security attack that aims to trick users in revealing personal information.Traditionally, email was the main means through which phishing attacks were executed. However,in recent years attention has shifted towards mobile messaging platforms like whatsapp.The aim of this project is to develop a mobile app that can be used to study how users behavetowards phishing and non-phishing mobile messages. The idea is to develop an app that will allowresearch to study mobile messaging phishing using experience sampling, a research methoddeveloped from psychologists to solicit user responses, feeling and attitudes of everydayexperiences. The main goal is to have a an Android app that will serve messages to users asnotifications including the message itself and visual previews of URLs, etc and offer options foruser to react to such notifications, e.g. dismiss, follow up, etc. User responses can also be followedby brief questions about capturing the user thinking. Behind the scenes, the app will record userreactions and question responses allowing researchers to collect data about user behaviours towardsmobile messages.The project requires strong programming skills and either existing knowledge of Androiddevelopment or undertaking of CS551 Mobile Software and Applications. The project also requiresthat you have an Android phone that you can use for development.TERZ07 – Copyright on BlockchainAcademic: Terzis, SotiriosDistributed ledgers and blockchains have received a lot of attention in recent years. One of theapplications suggested for the technology is to use to capture copyright information for producedmaterials.In this project the aim is to examine how easy it is to implement a blockchain system for copyright.The idea is to follow the the architecture suggested in “Towards an Open and Scalable MusicMetadata Layer” (https://arxiv.org/abs/1911.08278) and to see to what extent there are librariesavailable to support its various components and functions. Using the identified libraries the goalwill be to build a system bringing the various components together. A key challenge in this will beto identify the appropriate blockchain technology to use as a basis for the system. Note thatalthough the paper above is focusing on music works, the project doesn’t have to do so and could beinstead focus on book whose metadata are better understood.This project requires very strong development skills as it will involve getting to grips and using avariety of technical solutions.TERZ08 – Trust in Fog Computing SimulationAcademic: Terzis, SotiriosFog computing aims to address to weakness of cloud computing by introducing fog nodes close tothe edge devices (mobile phones, IoT devices, etc – referred to as end nodes) to support them withfunctions typically offloaded to the cloud, e.g. storage or computation. As end nodes rely on fognodes to carry out their functions, they need to trust them not to misbehave. Similarly, as fog nodesoften combine data from a number of end nodes, they need to trust them not to misbehave.Researchers have suggest various trust models to allow fog and end nodes to manage their trustrelationships. To study how these models perform researchers often use simulations.The aim of this project is to build simulator to study the fog computing trust model proposed in “Atwo-way trust management system for fog computing”(https://www.sciencedirect.com/science/article/pii/S0167739X19316437). The simulator should beable to run on Raspberry Pi’s. The idea is to use virtual machines running Raspberry Pi OS to playthe role of fog and end nodes in the simulation. These nodes will be able to simulate set of differenttrust related behaviours and will also run the trust management algorithm. The goal is to be able torun different configurations of behaviours and see whether the algorithm is able to identify misbehaving nodes.The project requires strong programming skills, while familiarity with Raspberry Pi programmingwill be beneficial.TERZ09 – Federated Learning StudyAcademic: Terzis, SotiriosFederated Learning (FL) is the idea of producing a learning model using data that are distributedbetween different parties typically not in an independent identically distributed way. FL hasreceived a lot of attention in recent years as it fits naturally with scenarios involving data producedby mobile or IoT devices.The aim of this project is to look at how different data distributions affect the quality of the learningmodels produced. The idea is to use TensorFlow Federated (https://www.tensorflow.org/federated)which provides already developed FL algorithms, like DP-FedAvg, and federated datasets that canbe used for experimentation. The main focus will be in producing different distributions of thedatasets including ones where certain data are not available, and to see how the resulting DPFedAvg model is affected.This project requires knowledge of Python and machine learning in general (e.g. taking CS985Machine Learning for Data Analytics).WALL01 – State machine generator for ArduinoAcademic: Wallace, WilliamState machines are a useful tool for structuring microcontroller projects. The objective of thisproject is to develop software to allow state machines to be drawn and then compiled to code thatcan be executed on microcontrollers using the Arduino development environment.This project potentially has a lot of code, so although it would be good to develop something thatworks end-to-end, the student is free to focus on one aspect of this, for example the “compiler” orthe UI.It is expected that this would be developed as an open source project that would be useful for theArduino community.WALL02 – Develop algorithm for Neo4j graph databaseAcademic: Wallace, WilliamNeo4j is the most popular graph database – by a large margin according to the DB-Engines Rankingof Graph DBMS. It is open source, with an Enterprise version and a freely available communityedition. The source code – Neo4j is written in Java – is available on GitHub.The supervisor is interested in extending Neo4j with a new algorithm, so it would be useful to lookat how this would be done and to document the process. The aim of this project is therefore todevelop an example algorithm that can be used to show how this is done, and to develop trainingmaterials, tests, etc to support this and ultimately contribute these in some way to the open sourcecommunity.Note that contributing to open source projects is a great way to sell yourself to potential employers.Google actively recruit by looking for “committers” to various open source repositories.WALL03 – Parallelise a graph algorithmAcademic: Wallace, WilliamIf you have an interest in learning more about parallel programming, including techniques such aslock-free data structures, then you’ll enjoy this project.The objective of the project is to take an existing algorithm and to modify it so that it works onmodern multi-core hardware. The new algorithm should be evaluated by comparing it against theserial implementation for speed and accuracy and by looking at how it scales with the number ofcores and threads.For evaluation of scalability, access will be given to a machine with 24 cores (48 threads).Although a parallel graph algorithm sounds complex, the methodology of this project is clear,standard data sets are available and the supervisor has a clear idea of the modifications needed tothe algorithm, so the plan for the dissertation is clear from day one.WALL04 – Speech Search for Movies and TV ShowsAcademic: Wallace, WilliamI have a Google Voice Kit sitting idle, and thought it would be interesting to look at using this as thefront end of a system used for searching for tv and movie content. The idea would be to carry outvoice to text analysis using one of the publicly available APIs online and then to carry out naturallanguage processing to figure out the intent of the speech and take some action.The action may “simply” be to use the speaker of the Voice Kit to tell the user what’s on, or perhapsit could use infra red to control a television or set-top box.There are several optional areas that could be looked at here, for example identifying the speakerfrom their “voiceprint” so that relevant recommendations could be made, or looking at privacyconcerns around the notion of recording speech that is then sent out onto the internet.In fact pretty much everything is optional except for the Voice Kit, so if you instead wanted tosuggest something more interesting like say a virtual companion for the elderly, I’d be happy tosupervise that.WALL05 – Brute Force Nearest Neighbour Search on Graphic CardsAcademic: Wallace, WilliamSystems such as recommendation engines often use nearest neighbour searches to recommendsimilar content. Finding the n-nearest neighbours is complex and is an area where there is on-goingresearch to do improve the speed and quality of the matching.Commodity graphics cards have literally thousands of processors and gigabytes of memory space.So, instead of trying to construct clever multi-dimensional indexes, we can instead use the bruteforce of such processors to search through every possible solution.The objective of this project is to develop a system to use CUDA (or OpenCL) to search sets of bothdense and sparse vectors and return the “top N” best matches.WALL06 – Control and monitoring software for prototype mixerAcademic: Wallace, WilliamA start-up company has a prototype mixer that mixes small batches of specialist cement. Theywould like to replace the current microcontroller and software with something that will allow themto carry out experiments to derive the best parameters for the mixing process.The aim would be to add some extra controls for motor speeds and timings and also add somesensors so that other parameters can be measured and recorded. As well as the software for themicrocontroller, a monitoring and control application that can be run on a laptop connected to themixer. This would allow the sensor readings to be recorded and (if time allows) visualised.Arduino devices would be suitable for this project due to easy availability of microcontrollers,hardware add-ons and software libraries.During this project, we may collaborate with colleagues in Chemistry and with partners in differentcompanies to look at how we could improve chemical production processes in different areas. Forexample one partner is looking at improving the production of plastic bottles to use a higherproportion of recycled PET plastic.WALL07 – Deep Neural Networks for Similarity Search of Cover Art ImagesAcademic: Wallace, WilliamIn a sense, Deep Neural Networks learn to encapsulate domain knowledge about the data set theyare being trained on. Typically the networks are trained as recognisers, but the knowledgeencapsulated can also be used to find similar data.The aim of this project is to build a demonstrator to show how images can be searched for startingwith images similar to those being sought.Although there are many datasets available for this, the supervisor would like if this was done formovie “cover art” images from a website such as “The Movie Database”.Another application of this is to look for images of counterfeit products on sites such as eBay andAmazon. A partner company is interested in looking at this.WALL08 – Voice Stress Analysis using Deep Neural NetworksAcademic: Wallace, WilliamDNNs used for speech-to-text pick up on attributes of the speech that can be used for differentapplications by applying transfer learning. For example voice stress analysis or identifying emotionin a voice.A recent review has highlighted new approaches to this problem using more traditional voiceanalysis tools:van Puyvelde, M., Neyt, X., McGlone, F., & Pattyn, N. (2018). Voice stress analysis: A newframework for voice and effort in human performance. Frontiers in Psychology, 9, 1-25.There are few data sets available in this area, but at least one has been identified. There is thepossibility to collaborate with colleagues in Psychology to design experiments to create a data setthat could be used for training.WALL09 – Combining signature files and word2vec to create meaningful,compact indexesAcademic: Wallace, WilliamSignature files are a method of indexing used for document retrieval that use less space thaninverted files. Whereas in inverted files you reserve a bit per word in the vocabulary, in signaturefiles, each bit is overloaded and hash functions are used on a word to select the bits.word2vec maps words onto a vector of numbers. If we threshold these to generate a vector of bits,we can generate a signature for a word where the bits have some semantic content.By combining these two concepts, we can build a compact index for documents that will mapsimilar documents together, even if they use different words.The supervisor would like to apply this to descriptions of movies and tv shows to see if it makesgood recommendations for similar content, but is happy for the student to use different data sets ifthey prefer.WALL10 – Natural language processing for the analysis of maintenance recordsAcademic: Wallace, WilliamPetrofac are a leading service provider to the international energy industry. With over 40 year’sexperience of providing a life of field service, which runs from concept to EPC, through tooperations and maintenance and decommissioning. As part of this service, Petrofac provide andoptimise maintenance solutions which relies on access to historic records and failure reports.Currently a significant volume of this maintenance data resides in databases with much of the detailrecorded in free text.The first part of this project is to apply the latest NLP techniques based on Deep Neural Networks(e.g. BERT, GPT-2 and ELMO) for entity extraction on these databases to identify items such aslocations, equipment being maintained, the fault, the fix, etc.By identifying these features it will become possible to report on these maintenance records andoptimise the maintenance process. For example, knowing that certain faults occur predominantly incertain regions might impact warehousing of spare parts, knowing the most common faults andfixes can minimise fault-finding time and help ensure that maintenance workers take the appropriateparts and materials along to a job and do not need to attend the same job twice. The second part ofthis project is to develop tools to allow this data to be explored.WALL11 – NLP for matching business problems to academic expertiseAcademic: Wallace, WilliamInterface (The knowledge connection for business) is a central hub connecting organisations from awide variety of national and international industries to all of Scotland’s universities, researchinstitutes and colleges.Interface works with businesses of all sizes, in all sectors, to match them to Scotland’s worldleading academic expertise to help them grow. In order to do this, they need to start from adescription of a business need and identify suitable academics that have the know-how to help withthis problem. The language and terminology used to describe business problems is different to thatused by academics to describe their research, so this matching is complex.The objective of this project is to apply the latest Natural Language Processing (NLP) techniques toa database of “Interface Enquiries” along with the academics who responded to them in order tobuild a smart matching system that can help Interface staff with this matching process. It isenvisaged that Deep Neural Networks (DNNs) such as BERT, ELMO and GPT-2 can be applied tothis task.Note that as an Information Retrieval task, this project would be suitable for an IM student.YANJ01 – A ransomware testbedAcademic: Yan, JeffRansomware is a pernicious form of malware that restricts an individual’s or enterprise’s access totheir device or data via strong encryption. In May 2021, a Russian cybercrime group shut down, viaa ransomware attack, the Colonial Pipeline in the USA, which carries refined gasoline and jet fuelfrom Texas to supply much of the East Coast. In response to this attack, the White House had todeclare a state of emergency.In this project, you will study common ransomware threats, and build a prototype of a testbed sothat ransomware can be examined and analysed in a restricted environment in a safe and secureway.Starting point:A Kharraz, S Arshad, C Mulliner, W Robertson, E Kirda, UNVEIL: A Large-Scale, AutomatedApproach to Detecting Ransomware, USENIX Security 2016. Available athttps://www.usenix.org/conference/usenixsecurity16/technical-sessions/presentation/kharazSupervisor: Jeff Yan (a new prof at CIS dept. whose research profile is available at https://profjeffyan.github.io/; please feel free to contact him at any time via [email protected])YANJ02 – Forensics for dronesAcademic: Yan, JeffDrones are no more merely for hobbyist leisure, but have found many mainstream applications. Inthis project, you will study common security and privacy threats facing drones, identify a set offorensic requirements for such devices, develop and evaluate a prototype of such forensic capabilitywhich you will design (either for real or via a simulation coded in your favourite programminglanguage).Supervisor: Jeff Yan (a new prof at CIS dept. whose research profile is available at https://profjeffyan.github.io/; please feel free to contact him at any time via [email protected])YANJ03 – Acoustic side-channels: novel attacksAcademic: Yan, JeffProject is suitable for 2 studentsSonarSnoop is an active acoustic side-channel attack which I conceived. Speakers are used to emithuman inaudible acoustic signals, and the echo is recorded via microphones, turning the acousticsystem of a smartphone into a sonar system. The echo signal can be used to profile user interactionwith the device. For example, a victim’s finger movements can be inferred to steal Android unlockpatterns. Details see the paper “SonarSnoop: active acoustic side-channel attacks. Int. J. Inf. Secur.19, 213–228 (2020)”In this project, you’re expected to extend SonarSnoop from the original setting of smart phones todifferent device types or a different application scenario. Many such generalisations were discussedin Section 6 of the above paper, and you are encouraged to choose your favourite option. All thesource code for the smartphone setting is available to play with and build on (in a responsiblemanner). Signal processing expertise is compulsory for this project.Supervisor: Jeff Yan (a new prof at CIS dept. whose research profile is available at https://profjeffyan.github.io/; please feel free to contact him at any time via [email protected])YANJ04 – Acoustic side-channels: countermeasuresAcademic: Yan, JeffProject is suitable for 2 studentsSonarSnoop is an active acoustic side-channel attack which I conceived. Speakers are used to emithuman inaudible acoustic signals, and the echo is recorded via microphones, turning the acousticsystem of a smartphone into a sonar system. The echo signal can be used to profile user interactionwith the device. For example, a victim’s finger movements can be inferred to steal Android unlockpatterns. Details see the paper “SonarSnoop: active acoustic side-channel attacks. Int. J. Inf. Secur.19, 213–228 (2020)”In this project, you’re expected to implement and evaluate empirical countermeasures againstSonarSnoop or other acoustic side-channel attacks. All the source code for SonarSnoop (thesmartphone setting) is available to play with and build on (in a responsible manner). Signalprocessing expertise is compulsory for this project.Supervisor: Jeff Yan (a new prof at CIS dept. whose research profile is available at https://profjeffyan.github.io/; please feel free to contact him at any time via [email protected])YANJ06 – Can TEEs help reduce user burden of solving CAPTCHAs?Academic: Yan, JeffProject is suitable for 2 studentsCAPTCHAs are the most common way for telling bots and humans apart on the Internet, but theycan be frustrating for human users due to usability and accessibility issues.In this project, you will explore a new approach that combines cryptographic protocols and clientside Trusted Execution Environments (TEEs) such as ARM TrustZone and Intel SGX, with the aimof reducing (or even removing entirely) user burden of solving Captchas.Intuitively, the key idea is about shifting some trust to crypto and enclave security.For those who are capable and like programming, this project can be a lot of fun, since you willneed get hands-on with web programming (both client and server), SGX programming, cryptoimplementations such as Merkle trees (which are fundamental for Bitcoins and blockchaintechnology) and security protocols.Starting point: “CACTI: Captcha Avoidance via Client-side TEE Integration,” in 30th USENIXSecurity Symposium, 2021.YANJ07 – Bespoke project 1 (in security, privacy, forensics or cybercrime)Academic: Yan, JeffWith a wide research interest in security, privacy, forensics and cybercrime, I am more than happyto work with students who’d like to propose your own ideas/projects in these areas. Or, we meet todesign a project together for you via discussions and brainstorming.Supervisor: Jeff Yan (a new prof at CIS dept. whose research profile is available at https://profjeffyan.github.io/; please feel free to contact him at any time via [email protected])YANJ08 – Bespoke project 2 (in security, privacy, forensics or cybercrime)Academic: Yan, JeffWith a wide research interest in security, privacy, forensics and cybercrime, I am more than happyto work with students who’d like to propose your own ideas/projects in these areas. Or, we meet todesign a project together for you via discussions and brainstorming.Supervisor: Jeff Yan (a new prof at CIS dept. whose research profile is available at https://profjeffyan.github.io/; please feel free to contact him at any time via [email protected])YANJ09 – Usable security for side-channel mitigationAcademic: Yan, JeffProject is suitable for 2 students“Hearing your touch” is an acoustic side-channel attack that recovers what users type on the virtualkeyboard of their touch-screen smartphone or tablet. When a user taps the screen with a finger, thetap generates a sound wave that propagates on the screen surface and in the air. We found thedevice’s microphone(s) can recover this wave and “hear” the finger’s touch, and the wave’sdistortions are characteristic of the tap’s location on the screen. Hence, by recording audio throughthe built-in microphone(s), a malicious app can infer PIN and text as the user enters it on theirdevice.Some application-level mechanisms appear useful to mitigate this “hearing your touch” attack, butthey raise usability concerns. This project is built on my recent paper:Ilia Shumailov, Laurent Simon, Jeff Yan, Ross Anderson. “Hearing your touch: A new acoustic sidechannel on smartphones”.https://arxiv.org/abs/1903.11137You will implement a few countermeasures as suggested in the paper, and empirically evaluate bothsecurity and usability of each countermeasure. Our ultimate goal is to find a countermeasure thatachieves good security and usability simultaneously, rather than at the expense of one another.All the source code for “Hearing your touch” is available to play with and build on (in a responsiblemanner).Supervisor: Jeff Yan (a new prof at CIS dept. whose research profile is available at https://profjeffyan.github.io/; please feel free to contact him at any time via [email protected])YANJ10 – Investigating gender bias in password managersAcademic: Yan, JeffProject is suitable for 2 studentsThis project investigates whether gender has an impact on the adoption and use of passwordmanagers, and how behavioural patterns in the use of password managers exhibit similarity anddifference for males and females. Relevant research methods include both qualitative andquantitative.No literature addresses this problem, and you will be one of the earliest researchers cracking it.However, there are a few papers on gender bias in cyber security, and about a dozen on the design,usability or adoption of password managers. It’s easy to identify them via Google Scholars.YANJ11 – Helping Dyslexic People with PasswordsAcademic: Yan, JeffDyslexic people struggle with alphanumeric passwords. This project aims to understand thesedifficulties (including their psychological processing roots), and to investigate how the difficultiescan be mitigated.This project will explore some novel designs based on solid theoretical backing. It involves withboth quantitative experiments and software development.YANJ12 – Does the Frequency of Severe Terrorist Events Evolve?Academic: Yan, JeffMore than 15 years ago, Aaron Clauset et al wrote an interesting and influential paper, “On theFrequency of Severe Terrorist Events”: https://arxiv.org/pdf/physics/0606007.pdfGiven that new data sets become available of terrorist attacks, a revisit of Clauset’s results is bothrelevant and interesting.In particular, I’m curious about the dynamics between the occurrence of terrorist attacks and thedrastically changed anti-terror policies. Have terrorists adjusted the occurrence pattern of theirattacks, in response to the US anti-terror policies adopted in the years after the 911 attack?We expect this project will inform govs, intelligence orgs and think tanks for policy purposes.Understanding conflicts on a macro-scale is intellectually interesting, too.A key part of this project is to (understand and) apply power laws and related mathematics/statisticsfor data analysis.YANJ13 – Investigating cybercrime and cyber criminalsAcademic: Yan, JeffProject is suitable for 2 studentsI have a number of ideas studying some specific cyber crimes, each with somehow differentprerequisites (e.g. whether you understand Chinese language or not; or whether you can program).For example, one candidate project is to try to work with online websites such as Gumtree,investigating rental scams/frauds. It involves with both experiments and programming.Email or chat, if you want more details.YANJ14 – Exploring learned Bloom filtersAcademic: Yan, JeffBloom filter is a probabilistic data structure for membership testing (i.e. whether an element is amember of a set). It offers a (significantly) compressed representation of a data set at the cost ofsome (configurable) false positives for membership query. Elegant and versatile, Bloom filter hasbeen deployed in numerous fielded applications including networking, security and big data.Recent research has discovered that machine learning can further improve the performance ofBloom filters. Thus, the so-called learned Bloom filters have attracted a lot of attention in bothacademia and industry. In this project, we will explore various features of learned Bloom filters, andlook for interesting applications (e.g. handling multidimensional data).Leading researchers in this area include Tim Kraska (MIT), Jeff Dean (Google) and MikeMitzenmacher (Harvard). However, the easiest starting points are:M. Mitzenmacher. A model for learned bloom filters and related structures. arXiv:1802.00884,2018.M. Mitzenmacher. Optimizing learned bloom filters by sandwiching. arXiv:1803.01474, 2018.Note: The concepts of both Bloom filters and the learned ones are very simple, but it requires somegood math and algorithm skills to look under the hood. If you have the right skill set, this projectcan be fun and useful experience. Otherwise, it might be painfully challenging.

• WE OFFER THE BEST CUSTOM PAPER WRITING SERVICES. WE HAVE DONE THIS QUESTION BEFORE, WE CAN ALSO DO IT FOR YOU.
• Assignment status: Already Solved By Our Experts
• (USA, AUS, UK & CA PhD. Writers)
• CLICK HERE TO GET A PROFESSIONAL WRITER TO WORK ON THIS PAPER AND OTHER SIMILAR PAPERS, GET A NON PLAGIARIZED PAPER FROM OUR EXPERTS

QUALITY: 100% ORIGINAL PAPER – NO PLAGIARISM – CUSTOM PAPER

Why Choose Us?

How It Works

• Click on the “Place Your Order” tab at the top menu or “Order Now” icon at the bottom and a new page will appear with an order form to be filled.
• Fill in your paper’s requirements in the “PAPER DETAILS” section.
• Fill in your paper’s academic level, deadline, and the required number of pages from the drop-down menus.
• Click “CREATE ACCOUNT & SIGN IN” to enter your registration details and get an account with us for record-keeping and then, click on “PROCEED TO CHECKOUT” at the bottom of the page.
• From there, the payment sections will show, follow the guided payment process and your order will be available for our writing team to work on it.

About AcademicWritersBay.com

AcademicWritersBay.com is an easy-to-use and reliable service that is ready to assist you with your papers 24/7/ 365days a year. 99% of our customers are happy with their papers. Our team is efficient and will always tackle your essay needs comprehensively assuring you of excellent results. Feel free to ask them anything concerning your essay demands or Order.

AcademicWritersBay.com is a private company that offers academic support and assistance to students at all levels. Our mission is to provide proficient and high quality academic services to our highly esteemed clients. AcademicWritersBay.com is equipped with competent and proficient writers to tackle all types of your academic needs, and provide you with excellent results. Most of our writers are holders of master’s degrees or PhDs, which is an surety of excellent results to our clients. We provide assistance to students all over the world.
We provide high quality term papers, research papers, essays, proposals, theses and many others. At AcademicWritersBay.com, you can be sure of excellent grades in your assignments and final exams.