School of Engineering and Informatics (for staff and students)

Junior Research Associates

Please note that the existing deadline of March 26th 2018 has been extended to April 16th 2018.

Junior Research Associates 2018

The University of Sussex Junior Research Associate (JRA) scheme is a pioneering project which aims to develop future research leaders.  It encourages talented and ambitious undergraduates to consider a career in research following graduation.  The Doctoral School offers a number of JRA bursaries worth up to £1,800 each. Students can either apply with one of the projects below or you can come up with an original proposal yourself (as long as a member of faculty is willing to supervise).

For further information about the scheme and how to apply visit the JRA webpages.

Applicants must be sponsored by a member of faculty so please contact the project supervisor to dicuss the project and application. See below for current projects within the Department of Informatics. Click on the project title to reveal a description:

Bio-inspired mobile devices for visual navigation

Project Supervisor: Prof. Andy Philippides

Currently, mobile phones rely principally on GPS for localisation, but the visual world also provides useful navigational information in situations where GPS is either unavailable (e.g. indoors) or where redundancy is necessary (e.g. for safety). Bees and ants use this visual information to achieve remarkable navigational performance despite small brains and low-resolution vision. Inspired by them, we have developed algorithms which exploit visual cues in an insect-like manner and can guide navigation along complex routes with minimal computation and memory [1-2]. These features make the algorithms suitable for use on mobile phones and we have developed a prototype app aimed at blind or visually impaired users [3]. We envision this JRA as building on this work, exploring how to best use visual information in a computationally efficient manner as well as comparing this with other modalities (i.e. WiFi and GPS data). However, we are open to other projects and for instance, an interested student could instead help build and develop an Android-powered robot platform capable of autonomous navigation [eg 4]. This project would suit a motivated student, with an interest in computer vision, robotics and/or app development. Experience with OpenCV or app development is useful but not essential.

[1] Philippides, A., Graham, P., Baddeley, B. and Husbands, P. (2015). Using neural networks to understand the information that guides behavior: a case study in visual navigation. Artificial Neural Networks, 227-244. Chicago.
[2] Baddeley B, Graham P, Philippides A & Husbands P (2012) A Model of Ant Route Navigation Driven by Scene Familiarity. PLoS Comput Biol 8(1): e1002336.
[3] Amos, D., Graham, P., & Philippides , A. (2016-17). Outcomes from a Sussex RDF project to test the viability of bio-inspired navigational devices.
[4] For example, see: https://blog.inf.ed.ac.uk/insectrobotics/roboant/

Deep learning for autonomous robot navigation

Project Supervisor: Prof. Andy Philippides

In this project, the JRA will employ cutting-edge deep learning methods to enable autonomous navigation on a wheeled robot equipped with an NVIDIA Jetson GPU. They will develop deep navigation algorithms (based on work developed in [1,2]) using Tensor Flow and deploy them on the robot using TensorRT. Results will be benchmarked against bio-mimetic approaches developed in the Brains on Board project and state-of-the-art navigation methods such as SLAM.

The project will require creative thinking and some programming experience. Experience with deep learning would be advantageous, but is not a strict requirement.

1. Philippides, A., Graham, P., Baddeley, B. and Husbands, P. (2015). Using neural networks to understand the information that guides behavior: a case study in visual navigation. Artificial Neural Networks, 227-244. Chicago.

2. Baddeley B, Graham P, Philippides A and Husbands P (2012) A Model of Ant Route Navigation Driven by Scene Familiarity. PLoS Comput Biol 8(1): e1002336 "

Designing an application for effective risk communication with fisherfolk

Project Supervisor: Dr Kate Howland

This project focuses on the design of an application that effectively communicates weather risk information to fishers in Thiruvananthapuram, Kerala. It contributes to the SSRP project ‘Co-production of knowledge and communication tools for safe and sustainable artisanal fishing’, which is combining local knowledge and scientific observations to better understand hazards, studying risk culture and testing innovative methods and tools. A range of communication platforms are being investigated, including mobile messaging, smart phone applications, loudspeakers, handheld internet devices, and VHF/ FM radio.

The JRA project will build on feedback from co-production workshops and communication technology tests carried out in collaboration with Indian National Centre for Ocean Information Services (INCOIS) and the State Disaster Management Agency (SDMA) of Kerala. The student will prototype a touch-screen based interface that conveys risk information clearly and effectively, in line with data from workshops, to run on a platform such as Android or Raspberry Pi.

For more information, contact Kate Howland k.l.howland@sussex.ac.uk

This project would suit a student with some interaction design experience and reasonable programming ability.

Slawson, Nicola (2017) Radio Monsoon aims to ensure safety reigns among fishermen in south India. The Guardian. https://www.theguardian.com/global-development/2017/apr/24/radio-monsoon-safety-fishermen-south-india-kerala

Prototyping conversational programming interfaces

Project Supervisor: Dr Kate Howland

VoiceUIs such as Amazon Echo and Google Home are increasingly used for smart home control, but provide little support for querying, debugging and customising the rules defining smart home behaviours through voice. Currently, these activities must be done using a separate, screen-based interface, as voice interaction is largely limited to triggering pre-defined behaviours. Automation platforms such as IFTTT allow programming of smart home behaviours through trigger-action rules, but have seen little uptake beyond early adopters and tech-savvy hobbyists.

The CONVER-SE project is investigating the challenges with using speech for programming through studies in real homes, and evaluating ways to mitigate these challenges, including conversational prompts, use of gesture and proximity data to avoid ambiguity, and providing default behaviours that can be customised.

The JRA project involves contributing to the design of an interface to support conversational programming in situ based on the findings from the in home studies. In particular, the student will develop dialogue paths to support querying and editing of existing rules, and elicitation of well-formed new rules.

For more information, contact Kate Howland k.l.howland@sussex.ac.uk

This project requires good programming skills and an interest in interaction design.

The CONVER-SE project: http://gow.epsrc.ac.uk/NGBOViewGrant.aspx?GrantRef=EP/R013993/1

Reinforcement learning and decision making in computational models of the insect brain

Main supervisor: Prof. Thomas Nowotny

Co-supervisor: Dr. James Bennett

Reinforcement learning algorithms enable an agent to learn associations between environmental cues and potential rewards, and thus learn to choose actions that are predicated on those cues so as to maximise cumulative rewards. We can learn a lot about reinforcement learning by looking to animals, which must learn to make decisions that increase the animal's chances of survival, e.g. where to look for food. In the Nowotny group, we are developing models of the insect brain that implements a well known reinforcement learning algorithm. There are a number of interesting questions to ask with these models. Examples include: how are decisions learned when multiple reward types e.g. food and water, must be obtained; how can reward reliability be learned and used in decision making? The exact research question can be discussed. Prerequisites: confidence with programming, and an interest in brains!

How much is your immune reaction determined by your genes? - Data Visualisation and Analysis

Project Supervisor: Dr. Bernhard Reus

A support tool for researchers in immunology is developed to visualise, analyse and compare patients’ amino-acid sequences of relevant alleles (variation of a gene). Based on patient data sets, the correlation between those amino-acid sequences and measurements related to patients’ health and immune reaction are to be investigated. The overall purpose of the tool is to understand how genetic disposition impacts immune reactions in patients.

This is a co-operation with the Immunology Department at BSMS. The JRA does not need to have any medical/biological knowledge, but should be a strong programmer with some interest in data visualisation and/or either statistics or machine learning.  

Lean - a Programming Language that doubles up as Theorem Prover

Project Supervisor: Dr. Bernhard Reus

Lean is a proof checker and a programming language in one. It is a collaborative open source project, hosted on GitHub (http://leanprover.github.io/) with involvement of Microsoft Research. A web browser embedded IDE is available at https://leanprover.github.io/programming_in_lean/?live . Lean’s reflective features allow one to program proof tactics.

After having gained familiarity with the Lean programming language by working through some examples, as provided in the tutorial https://leanprover.github.io/tutorial/tutorial.pdf, either strict equality types (i.e equality where all equality proofs are considered equal) or quotient (inductive) types shall be implemented in Lean and some proof tactics programmed for them.

Crime Locations in 18th and 19th Century London

Project Supervisor: Dr. Julie Weeds

The Old Bailey Online project provides a very rich source of information about court proceedings from 18th and 19th Century data, including both the transcripts of the trials and meta-data that was recorded at the time about the participants.  This project will seek to extract, analyse and ultimately map locations from the trials.  The student is expected to have good programming skills and a strong interest in natural language processing, including techniques for named entity recognition and relation extraction.  An interest in mapping and visualisation is also desirable.

Social Media and Mental Health

Project Supervisor: Dr. Julie Weeds

There are large quantities of text available from social media sites (e.g., internet forums and twitter) in which ordinary people discuss their experiences of mental health issues.  By collecting and analysing this data, this project seeks to offer insights into the commonly occurring co-morbidities i.e., the symptoms, conditions or disorders which commonly co-occur, particularly at the interface between mental and physical health.  One approach would be to automatically generate a thesaurus from the corpus data, which would relate commonly used terms based on their usage.  The student is expected to have good programming skills and a strong interest in natural language processing and machine learning.  In particular, interest in classification and clustering is desirable.

Cryptocurrency Transaction Visualisation

Project Supervisor: Dr. Martin White

In this project the JRA will investigate methods for visualising cryptocurrency, e.g. Bitcoin, Ethererum, network transactions flows with a view to identifying patterns that may indicate significant activities in the network such as: mining pools' operations; transaction patterns during high volatility periods in cryptocurrency exchange rates; wealth concentration in the network, etc. Furthermore, fraudulent or illicit activities may also be researched, such as the hacking events in the cryptocurrency ecosystem (e.g. the hacking of Mt. Gox exchange in 2014) or even directly in the blockchain (e.g. hacking of The DAO  on the Ethereum blockchain). This methodology was in part introduced at an early stage for the Topics in Computer Science module, and we wish to further develop it.

Cultural Blockchain

Project Supervisor: Dr. Martin White

In the this project the JRA will investigate how to develop a blockchain and associated smart contracts for a cultural blockchain. The idea is to develop a blockchain repository for museum collections catalog data with associated PREMIS metadata. The project will involve developing the cultural blockchain using Ethereum and writing smart contracts to access the museum catalog data for rendering in virtual museums scenarios.  The overall goal is to develop a proof of concept for the cultural blockchain.  We have recently developed blockchain and smart contract technology for a fake news application, which uses the same PREMIS metadata model for the blockchain data structure, read the paper here: http://online.liebertpub.com/doi/pdfplus/10.1089/big.2017.0071. We anticipate that the JRA will adapt our Fake News Blockchain to create the initial Cultural Blockchain.

Simulating Data Centre Transport Protocols

Project Supervisor: Dr. George Parisis

Data Centre Networks (DCNs) support the provision of core Internet services, such as search, social networking, cloud computing and video streaming. DCNs consist of a large number of commodity servers and switches,  supporting multiple paths among servers and very large aggregate bandwidth. TCP is ill-suited for meeting the throughput and latency requirements of applications in DCNs. During the last decade, there has been significant research on designing network protocols that enable efficient use of DCNs. This JRA is about the development of simulation models in Omnet++[1] for recently proposed DCN protocols, such as NDP [2] and MultiPath TCP [3], and the experimental evaluation of their performance using large-scale simulations.

[1] OMNeT++ Discrete Event Simulator, https://www.omnetpp.org/.

[2] M. Handley, C. Raiciu, A. Agache, A. Voinescu, A. W. Moore, G. Antichi, and M. Wójcik, Re-architecting datacenter networks and stacks for low latency and high performance. In Proceedings of ACM SIGCOMM '17.

[3] C. Raiciu, S. Barre, C. Pluntke, A. Greenhalgh, D. Wischik, and M. Handley, Improving datacenter performance and robustness with multipath TCP, In Proceedings of ACM SIGCOMM '11, 266-277.

Evaluating the Performance of Graph Processing Algorithms on Distributed Graph Processing Frameworks

Project Supervisor: Dr. George Parisis

Graph processing is a key aspect of big data processing. An extremely wide range of physical and non-physical constructs can be modelled as graphs and therefore efficient graph processing and calculation of local and global metrics is crucial for a large and diverse set of applications. Recently, a number of software frameworks for distributed graph processing have emerged (e.g. Pregel [1], GraphX [2], Giraph [3]). This JRA is about the development of widely used algorithms for graph characterisation (e.g. the calculating node centrality measures) in the most widely used distributed processing frameworks and the comparison of their performance using synthetic graphs.

[1] G. Malewicz, et al., Pregel: a system for large-scale graph processing, In Proc. of ACM SIGMOD, 2010.

[2] Xin RS et al., GraphX: a resilient distributed graph system on Spark, In Proc. of GRADES, 2013.

[3] Giraph, http://giraph.apache.org/

School of Engineering and Informatics (for staff and students)

School Office:
School of Engineering and Informatics, University of Sussex, Chichester 1 Room 002, Falmer, Brighton, BN1 9QJ
enquiries@enginf.sussex.ac.uk
T 01273 (67) 8195

School Office opening hours: Monday - Friday 09.00 - 17.00
School Office location [PDF 1.74MB]