Graham Kendall

Individual Projects (3rd Year Undergraduates)

(Academic Year 2008/09)


This page details the third year individual projects that are/have been supervised by Graham Kendall who is a Professor in the School of Computer Science at The University of Nottingham.

This page contains details of projects supervised in the academic year 2008/09.

The main project page can be found here.

Downloadable Dissertations

Some of the projects are available for download, so that you can view them. This facility is for current third year students, being supervised by me, who can use previous dissertations for their own research and also to look at previous dissertations to see how to lay them out etc. (it also saves the inevitable requests I receive asking to see a "good" project from last year, with the risk that I never see it again!).

As such, these files are password protected and, to get the username/password you need to EMAIL me. People outside of the university will not be given access.

In addition, I cannot give access to those students outside of my tutor group. You should approach your own tutor to get copies of previous dissertations they have supervised.

Note : I cannot make the files available in formats other than those already supplied.

3rd Year Projects Supervised in 2008/09

    1. Development of an Artificial Neural Network to Play OthelloAlexander by James Pinkney
    (pdf: 508KB)
    Abstract: The aims of this dissertation are twofold: firstly: to program an Othello engine which can be played by two human players, which will provide information to the players such as whose go it is, the current state of the board, and where legal moves can currently be made; and, secondly: to develop an Othello playing algorithm which has no prior knowledge of any tactics or strategies, which is capable of consistently beating human players and other artificial intelligences.
    The former is not necessarily a prerequisite for the latter, but will certainly be helpful for observing matches, and to act as referee, ensuring that the rules are enforced; for example, only legal moves should be allowed, and the board state should be correctly altered after a move has been made.
    It is proposed that the latter is to be developed by evolving a population of artificial neural networks which repeatedly play Othello matches against one another. After each network has played every other a certain number of times, the poorest players (determined by numbers of wins, losses and draws) will then be `killed off' and the strongest players duplicated and slightly modified, and so on, in effect, recursively generating better and better players, until such time as the system reaches a plateau.

2. Playlist Generation using a Hybrid of Recurrent Self-Organising Maps and Markov Models by Apurva Amin
(pdf: 1038KB)
Abstract:This study investigates the use of machine learning in monitoring user’s music listening habits and, subsequently, inferring the next song. Machine learning is an umbrella term for a subset of artificial techniques designed to allow computers to be able to simulate ‘learning’, not dissimilar to human brains. With large collections of music being available on user’s computers and media players, creating playlists can be time consuming, tedious and inaccurate. However,many current systems simply rely on a random function to generate playlists where each song is mutually exclusive from the next. This generally results in songs with no obvious patterns or connections, causing the user to have to manually change the playlist, either by skipping the song or directly selecting a new one. This dissertation analytically discusses current methods used to circumvent this problem, and proposes a new design which improves upon flaws found in the current paradigms. This work implements a hybrid of recurrent self-organizing maps, a form of unsupervised learning, and First order Markov chains in order to create a system which recognises habitual patterns and dynamically creates a playlist which it feels would be what the user would like to hear at that time. There are various upshots of the technique used which creates a much more feature rich experience for the user.

3. A Poker Player using Opponent Modelling on No-limit Texas Hold’em by Aadarsh Bhimarasetty
(pdf: 304KB)
Abstract:Poker is a popular card game with imperfect information. It has been a popular topic to create machines which play poker to play against human opponents and the recent work by the University of Alberta Computer Poker Research Group and The First Man-Machine Poker Competition in July 2007 has shown that it is feasible to create a poker playing robot to successfully play against human opponents.
This work shows a poker player using opponent modelling to adapt against opponents to win in games of Texas Hold'em Poker. This work is specifically suited to playing nolimit tournament games of poker against human opponents as well as other poker playing robots. The main feature of this work is opponent modelling.

4. Investigating the effects of machine learning to efficiently predict foreign exchange rates by Kushal Pisavadia
(pdf: 5,947KB)
Abstract:This document is the culmination of one years work as part of a final year assessed project for an E-Commerce & Digital Business Single Honours Degree. It discusses the artificial intelligence concept of “neural networks” and its effeciency in predicting foreign exchange rates. There are numerous constraints to the problem, which even if met may not produce an efficient solution due to external influences from the finance market. This report details areas such as the motivation for the project, research involved in completing the project, design of the FOREX prediction program, implementation of the program, reasoning for the techniques used, a discussion of the results obtained, success of research conducted and a discussion of potential project extensions.
A neural network system was created (using Java) and applied to three different problems: XOR, Sinusoid Waveform and FOREX rates. During training, five indicator inputs and a single target output were used on the network. The network was unable to efficiently predict raw time-series data, however it was successfully able to learn both the XOR function and to a lesser extent fit the sinusoid curve. The findings of the project confirm empirical studies on the usefulness of neural networks in raw time-series prediction.

5. Scheduling the Boxing Day and New Year’s Day Fixtures for the English Football League by Sam Rushton
(pdf: 1,190KB)
Abstract: This dissertation discusses the problem of scheduling fixtures for the English Football Leagues during the holiday period (Boxing Day and New Year‟s Day). Every year 92 teams must play a match on each one of these days, one being played at their home stadium, the other at their opponents. The main focus is to minimise the total distance travelled by all the teams; however there are a number of constraints which must be overcome to produce a feasible solution. These include controlling the number of teams playing at home on the same day that are in close proximity to each other. I define the problem and put forward a method which can be adapted to other similar problems. It uses a Depth First Search to produce a potential solution which can then be refined to a feasible solution. My results show significant improvements when compared to published fixture lists of previous seasons.

6. Martin Bailey
Abstract:Did not submit