06-06991 - Research Skills - MSc Student Presentations 2007/08

Tuesday 27th November 2007 - 15:00-16:30 - Sport+Exercise Science - LT3
Time	Speaker, Title and Abstract
15:00	BULLINARIA, John Introduction The module lecturer will remind everyone of the aims and objectives of these presentation sessions, and outline how they will be organized.
15:12	ALLSOPP, Christopher A Brief Introduction to EEG-Based Brain-Computer Interfaces Brain-computer interfaces (BCIs) can provide a link to the external world for those who lack the ability to use more traditional communications channels relying on motor skills. Electroencephalography (EEG) provides a non-invasive means of measuring the electrical activity of groups of neurons which can be analysed to detect the intent of users. This presentation seeks to introduce EEG-based brain-computer interfaces, providing an overview of relevant EEG characteristics and proposing techniques to detect the electrophysiological signatures of the intent of trained users.
15:24	CHAMBERLAIN, Oliver Fitness Functions for Art Art has existed for thousands of years, from caveman's early paintings through to modern day art such as Tracey Emin's 'My Bed'. For many decades people have been trying to build computer systems that can create art. There are many systems around that create evolutionary art and have done so with a fair degree of success. However most of these current systems rely on a user to judge the quality of the art to help guide the evolution. Surely this means the computer is still only a tool for the user who is still ultimately the artist? To have a computer which creates art means it must be able to judge the aesthetic quality of its own work and guide the evolution by itself w ith no outside interaction. Some work has been done in this area with varying degrees of success and their has been more success in the field of music, but can a computer ever truly judge a piece of art? Or will it always require a human to guide the evolution?
15:36	DANI, Shilpa Local Search and Distribution of Optima Local Search, also referred to as Neighborhood Search, is a widely used, general approach to solving hard optimization problems. In isolation, it is a simple iterative method for finding good approximate solutions - local optimum. A basic local search algorithm begins with an arbitrary solution and ends up in a local minimum where no further improvement is possible. Local optimum is the main problem with local search. Although these solutions may be of good quality, they are not necessarily optimal. Tabu search, belonging to the class of local search techniques, prevents local search from getting trapped in a local optimum, hence increasing the chances of hitting the global optimum. Tabu Search approach, which is based on a short-term memory (Tabu list), forces the algorithm to explore new areas of the search space. However, there is always the risk of cycling if all the escaping routes require constraint duration longer than the maximum number of iterations. The number of local optima and the distribution of the local optima in the search space therefore strongly influence the success of Tabu search. The distribution of optimal points in a search space affects different local search algorithms in different ways. This project will investigate this relation by studying a number of optimisation problems and local search algorithms. The well known 0-1 Knapsack Problem has been used for demonstrating the application of different local search techniques. Similar analysis would be conducted on other local search algorithms and optimisation problem at a later stage in the project.
15:48	DOWNS, Zachary Machine Layout in Factories Product manufacture in factories can be held back by inefficient floor-space management. Using nature-inspired or conventional techniques can help to optimise the layout of a factory floor, to improve the quality, productivity, and production time. Of particular interest is using evolutionary algorithms to tackle the problem.
16:00	MEESON, Alan Prediction of Signal Peptides and their cleavage positions using Fractal Prediction Machines Signal Peptides are sequences of typically between 15 and 30 amino acids found at the N-terminal part of a protein chain. These sequences act as a postcode for the protein, directing its transport within the cell. Once the protein has reached its destination the signal peptide is cleaved from the protein. The signal peptide is an important feature for a number of bioscience fields, such as drug design, and the prediction of signal peptides has been a subject of much research. A number of techniques have been applied to this problem including weight matrices, hidden Markov models and neural networks. This project explores the application of fractal prediction machines to the problem of predicting signal peptides and their cleavage positions.
16:12	WOOLFORD-LIM, Benjamin Evaluating Java Genetic Algorithm Libraries Genetic algorithms are a promising area of research in the field of search optimization. By using evolutionary techniques such as mutation, crossover, and selection based on fitness, effective solutions to a problem can be "evolved" over numerous iterations of an algorithm. As a result, toolkits have emerged to facilitate the implementation of GAs in various languages, including Java. This project aims to identify and evaluate several leading Java toolkits designed for that purpose, on such criteria as simplicity, usability, efficiency, and customisability.

Thursday 29th November 2006 - 14:00-15:00 - Computer Science - Room UG40
Time Speaker, Title and Abstract

14:00 PEAVOY, Daniel
A Study of Theoretical Aspects of Genetic Algorithms
[No abstract supplied]

14:12 ROBINSON, Carl
Investigation of Particle Swarm Optimization solutions to the Travelling Salesman Problem and possible real world application
In this presentation I will give an overview of the well known Travelling Salesman Problem (TSP), show the principles of Particle Swarm Optimisation (PSO) and explain an Evolutionary Computing (EC) technique that exists for TSP. I will then explain what I intend to show in my mini project and answer questions on my project topic.

14:24 SIMPSON, David
Invariance and Neural Networks
In humans, classification of an object is almost always invariant to its size, position and rotation despite the fact that even small transpositions of the object can result in massive changes in the retinal input. When building visual systems using neural networks it is important to include this invariance in any practical system. In this presentation we will see how Humans and simpler organisms obtain invariant recognition of objects. Solutions for neural network systems will also be proposed along with an example of how translational, scale and rotational invariance has been achieved in one particular face recognition system.

14:36 STEAD, Richard
Real-time Musical Instrument Recognition
Identification of musical instruments active in a live ensemble is desirable, with educational value to musicians, in relation to automatic transcription and training in orchestration. Such automatic recognition of multiple signal sources has not yet been demonstrated satisfactorily. Research has focused on single instrument recognition based on various approaches of objectively describing the sound qualities of musical instruments. An overview of the area is presented together with a initial approach to solving the classification problem.

Tuesday 4th December 2007 - 15:00-16:00 - Sport+Exercise Science - LT3
Time Speaker, Title and Abstract

15:00 WIRTH, Christopher
Investigating Musical Improvisation Through Nature Inspired Methods
Many evolutionary programs have been developed that attempt to give computers the power to compose and play music. One such program is GenJam, which also has a feature allowing it to improvise by listening to a human play and then playing back an appropriate response.
This presentation addresses the underlying structure of GenJam and looks at the way in which the program is able to improvise with a human, before going on to explain the practical work that I will be doing to extend this feature further.

15:12 CROWTHER, Daniel
3D Illusionary Software Research
3D modeling is a method of attempting to represent a 3-Dimensional environment, known as a Universe, through a 2- Dimensional medium such as a computer screen. This project is intended to research the feasibility of applying illusionary imaging techniques to existing 3D modelers in order to produce real-time 3-Dimensional illusions.

Thursday 6th December 2006 - 14:00-15:00 - Computer Science - Room UG40
Time Speaker, Title and Abstract

14:00 GARCIA, Adrian
Genetic Algorithms and Stock Cutting
Genetic algorithms (GAs) are approaches to computing that are based on the ideas of evolution from the biological sciences. Stock cutting is a class of problem commonly found in industry, where fixed-sized stock of some material needs to be cut in a way that minimises waste. This presentation will outline some motivations for using a GA approach to solving a stock cutting problem and will explore some of the technical issues involved in implementing a GA.

14:12 GHARIB, Tony
Mobile Ad Hoc Protocols
A mobile Ad Hoc Network is a collection of mobile nodes connected without the required intervention of any access point. The nodes are free to move randomly and organise themselves arbitrarily. Such a network may or may not be connected to the Internet. Since the nodes do not have a fixed infrastructure there need to be a routing protocol to discover routes between the nodes. This presentation will introduce the concepts of Mobile Ad Hoc Networks and discuss some of the available routing protocols.

14:24 JONES, Llion
Integer Sequence Prediction via Hypothesis Segregation
In this presentation I will describe the problem of Integer Sequence Prediction; given a sequence of integers, we can ask what the next number is, or even how does this sequence continue. The background of this problem and the current solutions are discussed, before introducing a novel approach under the title of 'Hypothesis Segregation'. The key ideas of Hypothesis Segregation are demonstrated before glossing over some implementation issues. I conclude by comparing features of the current solutions to my proposed solution, such as what class of Integer Sequences they are able to successfully analyse.

14:36 PLESIAK, Wojciech
FPGA Design
Hardware design is a very difficult and time consuming process. Containing several stages, including coding, simulating and physical mapping, it differs much from software design process. Using widespread software programming languages, instead of heavy Hardware Description Languages, would save time and designer's effort and would improve error tracking.
An overview of Field Programmable Gate Array, its structure and application, is given. Differences between software and hardware programming are briefly discussed. A description of programmable logic design process is presented. Finally, the new approach for hardware programming is introduced.

Tuesday 11th December 2007 - 15:00-16:00 - Sport+Exercise Science - LT3
Time Speaker, Title and Abstract

15:00 ROBINSON, Tristan
Augmenting Situational Awareness using Mobile Computing
Situational awareness is the continuous extraction of environmental information. People with severe visual impairment need a means of remaining oriented to their environment as they move through it. While current mobility devices such as the cane and guide dog serve their purpose, they do not provide macro-navigation to their user.
This has led me to develop a specification for a wearable orientation device that helps blind users improve their situational awareness by providing information on buildings within their local environment. The mobile travel aid aims to increase the independent mobility of blind travellers, and builds upon current Bluetooth technologies to interact with nearby buildings.

15:12 RUPRAI, Manpreet
Cost Estimation for the Aerospace Industry
Cost estimation is an issue which concerns a number of industries as it will support planning and control for the future, and for aerospace industries it is explicitly important as new projects can cost millions and the viability of a project needs to be conclusive before a project is commenced. The development time of an aerospace project can span over many years, hence data from previous projects is limited, and this makes the task of producing a justifiable estimate ever more challenging. This project will attempt to study possible mathematical and computational methods to produce a justifiable estimate for the aerospace industry.

15:24 SABEGHI SAROUI, Behrang
Cryptanalysis and RC5
Cryptography is concerned with analysis of ciphers to discover methods to obtain the meaning of encrypted information. In this presentation I will give a brief introduction on Cryptanalysis and Cryptography. After that I will be talking about a cipher called RC5 and possible cryptanalysis techniques to attack this particular cipher.

15:36 SINGH BIRAN, Bobby
Mobile P2P
In recent years the growth of P2P networks has seen unprecedented growth. With the growth of mobile computing the P2P concept has been introduced into this context. The aim of the research was to gain a understanding of the current research field and to assess the feasibility of a Mobile P2P system.

Thursday 13th December 2006 - 14:00-15:00 - Computer Science - Room UG40
Time Speaker, Title and Abstract

14:00 TALBOT, Maxwell
Investigating the Scalability of Multi-Objective Evolutionary Algorithms
Multi-objective evolutionary algorithms (MOEAs) have received a lot of attention in the past, in the hope that they would enable the production of a diverse set of Pareto-optimal solutions in a single simulation run. There exists a number of different algorithms, most notably NSGA-II, SPEA2 and PAES, which have been created specifically to obtain such solutions efficiently. However, testing of these algorithms has been limited generally to the use of two or three objectives. In order to understand the algorithms and, more importantly, the limits to their applications better, the effects of adding more objectives as well as more decision variables will need to be studied more closely.

14:12 DEVLIN, Hayden
The Credibility of Evolutionary Artworks
In 1987 Richard Dawkins publicised his critically acclaimed "The Blind Watchmaker" in which he presented the first evolutionary image generation system with his famous 'Biomorphs' program. Since then the process of creating imagery through evolutionary techniques has become a field of natural computation in itself.
The method has been adopted by artists and scientists alike and its artistic potential will be discussed, looking on examples of 'evolutionary artists', and critiquing their approaches. Finally possibilities to improve the artistic application of the technique will be considered.

14:24 YAMAMOTO, Morihito
Reference Extractor
Since the popularization of the internet, a style of information retrieving in a research seemed to have been changed by the development of the online services. There are a great number of online journal articles, papers, and the other academic resources, which can be published at institutions' level, publishers' level, and even a personal level. Because of this advantage of the resources, these online academic resources often enable researchers to access and find up-to-date information. However, online resources is not always organised and it might be difficult for researchers to collect the papers related each other in terms of a specific topic. Thus, reference linking has been obtaining a great attention and developed by many researchers to suggest efficient information retrieving style. The reference linking is to connect related research by association with reference sections of each academic paper. As one of the core issue of reference linking, reference extraction can be introduced. Reference extractor extracts a reference into metadata, such as authors' name, title, year of publication, to recognize related studies. Moreover, there appears to be many different approaches towards reference extraction. Therefore, in this research, each approach towards reference extraction will be compared and a reference extractor will be programmed referring to the approaches.

This page is maintained by John Bullinaria. Last updated on 8 January 2007.

Thursday 29th November 2006 - 14:00-15:00 - Computer Science - Room UG40
Time	Speaker, Title and Abstract
14:00	PEAVOY, Daniel A Study of Theoretical Aspects of Genetic Algorithms [No abstract supplied]
14:12	ROBINSON, Carl Investigation of Particle Swarm Optimization solutions to the Travelling Salesman Problem and possible real world application In this presentation I will give an overview of the well known Travelling Salesman Problem (TSP), show the principles of Particle Swarm Optimisation (PSO) and explain an Evolutionary Computing (EC) technique that exists for TSP. I will then explain what I intend to show in my mini project and answer questions on my project topic.
14:24	SIMPSON, David Invariance and Neural Networks In humans, classification of an object is almost always invariant to its size, position and rotation despite the fact that even small transpositions of the object can result in massive changes in the retinal input. When building visual systems using neural networks it is important to include this invariance in any practical system. In this presentation we will see how Humans and simpler organisms obtain invariant recognition of objects. Solutions for neural network systems will also be proposed along with an example of how translational, scale and rotational invariance has been achieved in one particular face recognition system.
14:36	STEAD, Richard Real-time Musical Instrument Recognition Identification of musical instruments active in a live ensemble is desirable, with educational value to musicians, in relation to automatic transcription and training in orchestration. Such automatic recognition of multiple signal sources has not yet been demonstrated satisfactorily. Research has focused on single instrument recognition based on various approaches of objectively describing the sound qualities of musical instruments. An overview of the area is presented together with a initial approach to solving the classification problem.