RA Symposium 2018

Welcome to the 2018 Department of Computing Research Associate Symposium

The DoC RA Symposium is a one-day workshop to showcase the wide range of interesting computing-related research undertaken by postdoctoral research associates and fellows in the College to other research staff and students of DoC and other departments. As we aimed to celebrate your work, all RAs were invited to present their work at the symposium, as long as it’s computing-related.

THE WORKSHOP WAS A SUCCESS, WE HAD GREAT TALKS AND DISCUSSIONS

• Date & Time: 9:30 – 17:30 on Friday 22 June 2018
• Venue: Lecture Theatre 311, Huxley Building, Department of Computing, Imperial College South Kensington Campus, 180 Queen’s Gate, SW7 2AZ London

Programme

Schedule

Keynote

We were proud to have Dr Ruth Misener as the keynote speaker. Ruth talked about computational optimisation and its applications to various fields of engineering.

Ruth Misener is a Senior Lecturer in the Computational Optimisation Group, DoC, Imperial. Foundations of her research are in numerical optimisation algorithms and computational software frameworks. Applications include bioprocess optimisation under uncertainty and petrochemical process network design and operations.

Ruth holds an EPSRC Early Career Fellowship (2017-2022). Ruth received the Sir George Macfarlane Medal for excellence in the early stage of her career, as the overall winner of the 2017 RAEng Engineers Trust Young Engineer of the Year competition among young engineers who have made a major impact in their respective fields early in their careers.

Talks

We had 7 talks, as follows.

1. Paul-Antoine Arras (DoC): Selective binary rewriting
   Binary rewriting is a technique that consists in disassembling a program in order to add, remove or replace some or all of its instructions. It is used in a broad range of applications: sandboxing, code hardening, fault injection, tracing, fuzzing, multi-version execution, etc. However powerful and versatile, this scheme is known to be extremely hard to apply properly and usually results in heavy implementations. Common options include statically rewriting the binary file on disk prior to execution, as well as dynamically translating the assembly as and when it is run. Both have well-known shortcomings in terms of soundness and performance.
   This talk aims at first presenting an overview of these options with their respective benefits and drawbacks. Second, it will explain the fundamental transverse challenges of binary rewriting. Finally, it will depict a novel approach that has the potential to outperform existing schemes for most practical applications.
   Keywords: Operating Systems, Hardware/Software Interface, Low-Level Programming.
2. Erisa Karafili (DoC): Analyzing cyber-forensics evidence: A formal approach
   The frequency and harmfulness of cyber-attacks are increasing every day, and with them also the amount of data that the cyber-forensics analysts need to collect and analyze. We propose a formal analysis process that allows an analyst to filter the enormous amount of evidence collected and either identify crucial information about the attack (e.g., when it occurred, its culprit, its target) or, at the very least, perform a pre-analysis to reduce the complexity of the problem in order to then draw conclusions more swiftly and efficiently. Our main contribution in this work is the introduction of the Evidence Logic (EL), which allows an analyst to represent the different pieces of evidence, together with their sources and relations of trust, and reason about them, by eliminating the conflicting pieces of evidence during the analysis process.
   The Evidence Logic is based on Linear Temporal Logic and represents simple and derived pieces of evidence from different sources. We propose a procedure, based on monotonic reasoning, that rewrites the pieces of evidence with the use of tableau rules, based on relations of trust between sources and the reasoning behind the derived evidence, and gives as result a consistent set of pieces of evidence. The procedure is the EL rewriting system which, given pieces of evidence, yields a consistent set of pieces of evidence by rewriting the pieces of evidences and resolving their discordances. As proof of concept, we apply our analysis process to a concrete cyber-forensics case study.
   Keywords: Cyber Attacks, Formal Methods, Digital Forensics, Linear Temporal Logic, Tableau Rules.
3. John Cupitt (DoC): libvips – A fast image processing library with low memory needs
   Digital images are used everywhere — in our phones, on the web, and in almost every branch of research, from medicine to archaeology. If we can process these images more quickly, we can potentially accelerate many fields of study. Most image processing libraries operate on complete images. When performing a series of operations on an image of more than a few megabytes, pixels must be repeatedly fetched from main memory, operated upon, and written back. Developed at Imperial over the last 10 years, the libvips image processing library can compute complex sequences of operations on very large images without going via main memory, and can spread work over available CPU cores with almost no locking. This talk will present the architecture of libvips, show benchmarks, and illustrate its use with applications in medical imaging, archaeology and on the web.
   Keywords: Digital Image, Medical Imaging, Archaeological Imaging.
4. Sajad Saeedi Gharahbolagh (DoC): Dynamic SLAM
   In computer vision and robotics community, Simultaneous Localisation and Mapping (SLAM) is a well-known problem. Using SLAM, a sensor, such as a camera, is able to localise itself in an unknown environment by incrementally building a map and at the same time localising itself within the map. Various methods have been proposed to solve the SLAM problem, but robustness and real-time performance is still challenging. Dynamic environments are challenging for SLAM algorithms, since they cause wrong estimates and also introduce corrupted objects to the model. Specially large dynamic objects such as a moving bus or truck can easily misguide the algorithm. Thus it is important to know which objects are dynamic and which ones are static. We develop a dynamic SLAM algorithm which is able to identify dynamic object and exclude them from the pose estimation algorithm. We provide a dynamic map, where a static world is shown along trajectories for dynamic objects.
   Keywords: SLAM, Moving Objects, Neural Networks.
5. Deborah Adkins (Dyson School of Design Engineering): Decision making for energy efficiency in the domestic built environment
   In order to meet current climate change objectives, a massive overhaul of the existing housing stock is required. People would like to retrofit their homes with the latest eco-interventions but when faced with a slew of inconsistent information and marketing materials, the decision on what technologies and solutions to implement in a given application represents a significant challenge and the means to come to such decisions is the focus of this talk.
   This talk introduces a novel tangible object called HOUSE (Home User and Stakeholder Environment) to facilitate retrofit selection. Deborah will talk about the findings of a participatory modelling study that investigated the impact of HOUSE on dialogue and decision-making. She will discuss how a boundary object approach to eco-design can enable improved stakeholder interaction and how the tactile nature of the tool enables wider participation and engagement.
   Keywords: Built Environment, Design, Decision Making.
6. Liang Yang (Department of Earth Science and Engineering): PyEFEM: Highly parallel python based framework for flow and fluid-structure interaction (FSI) problem
   Numerical simulations of complex problems, such as turbulent flows, which demand fine spatial and temporal resolutions, require often impractical (or unavailable) computing resources. Current trends in supercomputing include the reduction of the cost of memory per core, achieving better computational performance and portability of the software, including on GPUs. For example, the explicit variational multiscale method (EVMS) is a promising algorithm for solving the incompressible, turbulent flow problems, with reduced memory costs. Implementing EVMS, PyEFEM is a Python based framework for solving flow problems using matrix-free stabilised finite element, which is capable of simulating large scale flow problem involved 10 billion unknowns on HPC. It is designed for massively-parallel multi-platform and its applications include: aerodynamics; biological flow, aeroacoustics, porous media flow, environmental flow and haemodynamics.
   (The EVMS method appears as front cover to the recent edition of Computational Mechanics [1].)
   [1] Yang, L., Badia, S., & Codina, R. (2016). A pseudo-compressible variational multiscale solver for turbulent incompressible flows. Computational mechanics, 58(6), 1051-1069.
   Keywords: Stabilised Finite Element Method, Computational Fluid Dynamics, High Performance Computing.
7. Tsipa Argyro (Department of Bioengineering): Design of optimal bio-processes through a hybrid modelling framework: A case study in bioremediation
   Currently, design and optimisation of biotechnological bioprocesses is performed either through exhaustive experimentation and/or with the use of empirical, unstructured growth kinetics models. Whereas elaborate systems biology approaches have been recently explored, studies of microorganisms fed by multiple substrates are predominantly ignored despite their significance in enhancing bioprocess’ performance. Herein, bioprocess optimisation for an industrially-relevant bioprocess involving a mixture of highly toxic substrates was achieved through application of a novel experimental-modelling gene regulatory network – growth kinetic (GRN-GK) hybrid framework. The GRN model described targeted gene regulatory networks and informed the formulation of the growth kinetics model replacing the empirical and unstructured Monod kinetics. The GRN-GK framework’s predictive capability and potential as a systematic optimal bioprocess design tool, was demonstrated by effectively predicting bioprocess’ performance, which was in agreement with experimental values, when compared to four commonly used models that deviated significantly from the experimental values. Significantly, strong evidence of model-based bioprocess optimisation was achieved, rendering the GRN-GK framework as a novel and applicable approach to optimal bioprocess design. Finally, model analysis using global sensitivity analysis (GSA) suggests an alternative, systematic approach for model-driven optimisation in synthetic biology and metabolic engineering applications.
   Keywords: Bioprocess Optimisation, Model-Based Control, Global Sensitivity Analysis.

Discussion Panel

We had a discussion panel On Career Perspectives, i.e. What to do after RA at Imperial?, lasting over an hour, with guest panellists:

• Mohammed Elmikaty – Lead Engineer at Jaguar Land Rover
• Erisa Karafili – Marie Curie Individual Fellow at DoC, Imperial
• Giacomo Tarroni – Medical Imaging Research Fellow at DoC, Imperial
• Ruth Misener – Senior Lecturer at DoC, Imperial

Call for Papers

Abstract submission is now CLOSED.

We welcome contributions, theoretical and practical, from any area of computer science. We also greatly encourage contributions from any other scientific discipline that could benefit from and build on state-of-the-art computing methodologies to define and solve new research questions.

We are open to all scientific disciplines and welcome researchers from all departments, including (but not limited to) Medicine, Natural Sciences and Engineering. We invite RAs to share their problems, solutions and ideas to foster collaborations both among groups and between departments.

If you would like to present your work in this symposium, please submit your abstract (100 to 500 words) to DoC RA reps (doc-ra-reps@imperial.ac.uk) by Tuesday 22th May 2018. The format of the talks (20+5 minutes) is semi-formal and there will not be formal proceedings; work-in-progress talks are especially welcome. All submissions will receive feedback from us before the symposium.