| Tuesday 15th January 2008 - 14:00-16:00 - Learning Centre - UG06 | |
|---|---|
| Time | Speaker, Title and Abstract |
| 14:00 |
John Bullinaria
Introduction The module lecturer will remind everyone of the aims and objectives of these presentation sessions, and outline how they will be organized.
|
| 14:10 |
Horia Corcalciuc
Secure Programming I shall be giving a speech about the current and most common deficiencies in programming languages. An attempt will be made to sum up the problems arising from traditional programming and their implications in computer security. Two case studies shall be presented in order to demonstrate the need for "logical extensions" to current reasoning in programming languages. Some current solutions and approaches will be debated pointing out their inability to solve problems in advanced programming topics such as concurrency. The speech will try to give a small but consistent overview of possible directions in reasoning about security at language level.
|
| 14:30 |
Mohamed Menaa
Runtime Visualisation Using Aspect-Oriented Programming Runtime visualisation is a powerful, yet undervalued medium that can help manage the increasing complexity in software. It can be used for debugging, testing, profiling, and optimising software. Moreover, it proved to be a significant pedagogical tool. Seemingly, the reason behind its neglect is the difficulty of instrumenting programs for runtime information without introducing errors and large overheads. Aspect-oriented programming (AOP) offers a simple, elegant, and efficient solution to this problem. In this talk, I discuss the problems posed by some current approaches to program instrumentation for runtime visualisation, then introduce the aspect-oriented alternative. AspectJ is used to epitomise a few recurrent schemes. Time permitting, I will briefly demonstrate my proof-of-concept tool "Ara", a generic runtime visualisation tool for Java programs.
|
| 14:50 |
Victor Landassuri-Moreno
Modular Support Vector Machines for Time Series Forecasting Until now we have a great diversity of algorithms that let us solve diverse problems with Neural Networks in an accurate way, and the possibility of have a method to combine some them in a structure to be able to solve more than one problem sounds tantalizing. In this sense, this research will be based in the development of a meta-structure (a modified Modular Support Vector Machine) that let us solve several problems on the same field, in this case, the regression problem for Time Series forecasting. This algorithm will have two kinds of modules: a) modules able to solve more than one problem of the same class and b) modules focus to solve a unique task. It is expected to have modules able to collaborate among them to have a better solution or to solve another task of the same class. Then it will let us have a compact and incremental architecture. Better results are expected based in the improvement of Support Vector Machines over classical Neural Networks, i.e. the ill-posed problem.
|
| 15:10 |
Jonathan Edwards (MSc student)
Improving the Evolvability of Combinatorial Logic Circuits with Pseudo-Quantum Logic Gates The evolution of combinatorial logic circuits is made difficult by the existence of large plateaus in the fitness landscape. Following an overview of combinatorial circuits, the standards for evolving them (representation, fitness, selection, etc), and a few specific algorithm as examples of research into this area, the field of quantum inspired algorithms is introduced, before a new algorithm for the evolution of combinatorial logic circuits is presented. By utilizing pseudo-quantum gates that are a super position of multiple conventional gates, a pseudo-quantum circuit represents a class of conventional circuits, and hence can use more global information about the search space from beyond the current plateau, to drive evolution.
|
| Friday 18th January 2008 - 14:00-16:00 - Learning Centre - LG33 | |
|---|---|
| Time | Speaker, Title and Abstract |
| 14:00 |
John Bullinaria
Introduction The module lecturer will remind everyone of the aims and objectives of these presentation sessions, and outline how they will be organized.
|
| 14:10 |
Roly Perera
Programming Languages for Interactive Computing Traditional programming languages are algorithmic: they are best suited to writing programs that acquire all their inputs before executing and only produce a result on termination. By contrast most applications are interactive, in that they maintain ongoing interactions with their environments. Modern real-world systems address this incompatibility by embedding an event-driven state management scheme into the host language which dispatches fragments of imperative code in response to external events, the job of each executed fragment being to restore the internal consistency of the computation. The downside of this approach to interaction is that it relies heavily on mutable stores and side-effects and mixes application logic in with behaviour which is arguably the responsibility of an execution model. I propose an alternative programming model which supports interaction directly. The distinguishing feature of the model is its modal construal of state and interaction.
|
| 14:30 |
Amaria Zidouk
A Model of Light Propagation in the Tissues of the Oesophagus Cancer of the oesophagus (gullet) has become the ninth most common cancer in adults in the United Kingdom. Unfortunately, early malignant tumours in the gullet cannot be seen with the naked eye and detection relies on conducting biopsies after the observation of abnormalities in the lining of the oesophagus. The random nature of this process implies that some early cancers may not be detected. This talk outlines a proposal for creating a physically realistic model of light propagation in the tissues of the oesophagus, which can later aid in the detection of gullet cancer. To this end, we will present the models used and describe some of the tools that might assist in the research process.
|
| 14:50 |
Alex Shaw
Separation Logic and Security A look at how Hoare logic and separation logic can be applied to security.
|
| 15:10 |
Shuo Wang
Churn Prediction CRM (Customer Relationship Management) is a popular topic in recent years. Especially in the fiercely competitive telecommunications market, it presents a great many challenges to providers of these services, but at the same time affords extraordinary opportunities to those providers who understand the market and, most importantly, the needs and wants of existing and potential customers. How to apply technical methods to help decision-making process and gain more profit during business is more and more important. Churn Prediction is a typical CRM problem that many tele-companies are facing today. In this talk, I will introduce what churn prediction is and why we need it. Then, a common framework and popular modeling and optimization solutions will be discussed.
|
| 15:30 |
Noureddin Sadawi
Human Face Detection and Recognition Human face detection in still images has been researched extensively over the past decade, due to the recent emergence of applications such as security access control, visual surveillance and content-based information retrieval. How-ever, this problem remains challenging because faces are non-rigid objects that have a high degree of variability in size, shape, color and texture. In this presentation, I am going to introduce the major efforts carried out in this cumbersome research field. The presentation will outline the advantages and disadvantages of the different techniques used, the challenges faced, and the futures trends.
|