NSC home > group seminars

Non-Standard Computation Group

Group seminars list of speakers and abstracts: 2008-09


These seminars are run jointly between the Non-Standard Computation Group in Computer Science and the Intelligent Systems Group in Electronics.

Speaker Title and Abstract Date and Location
Yang Lui (Elec) A Novel Low-level Image Segmentation and Tracking Technique Using Evolutionary Cellular Neuro-Immune Network.
In this talk a novel image processing system Cellular Neuro-Immune Network is proposed which is inspired by biological neuro-immune interaction. This image processing technique enhances the respective ability of image segmentation and edge detection. We make use of a Compact Genetic Algorithm to tune neuro-immune ``binding'' forces to achieve both robustness and fastness abilities. This technique is effective especially in a circumstance where very limited computation resources are provided e.g. FPGA's, and also effective when applied into real time visual tracking system and driven by other off-the-shelf techniques.
1315-1415 24th Oct, 2008 G/020
Fiona Polack (CS) Simulation, Validation and Complex Systems:Complex systems are typically modelled by simulation. Here, simulation means computer modelling, typically with a visualisation; it is attractive because it allows the essential temporal dimension of complexity to be incorporated in a model. In computer science, many such simulations are inspired hacks: knowing whether and to what extent the simulation model represents anything realistic is a long-running concern. The CoSMoS project, in seeking to support complex systems simulation, has been looking at simulation and validation. This talk presents some early research in this area, and is a reworking of a talk given by Paul Andrews and Fiona Polack at the first CoSMoS workshop, in September. 1315-1415 7th Nov, 2008 CS103
Ed Clark and Simon Hickenbotham (CS) Plazzmid: getting more bio into bioinspired evolutionary algorithms: The goal of the Plazzmid project is to develop a novel flexible and extensible computational framework and toolset inspired by sophisticated models of complex biological evolutionary processes that occur in bacteria (and in bees). The resulting tools will be used both to build and analyse testable models of biological evolutionary processes, and to build and analyse powerful novel computational metaphors and algorithms based on these more sophisticated biological models. The efficiency and adaptability of the bacterial genome make it a good model for evolutionary algorithms, but it is tightly coupled with the components of the bacterial metabololism. This talk gives an overview of the features of the bacterial genome we take inspiration from. It explores an approach to modeling the bacterial metabolome in an efficient and modular manner, so that bacterial genome organisation and gene regulation can be implemented in evolutionary algorithms. The potential of this model is demonstrated by creating a network that is capable of appropriate behavioural switching that can be observed in bacteria. 1315-1415 21st Nov, 2008 CS103
Matt Speed (Elec) Finite Difference Techniques for Voice Synthesis: The increased availability and accuracy of MRI data describing the human vocal tract affords the development of a new generation of flexible voice synthesis systems. This seminar explores the application of finite-difference approximation techniques to the problem of voice synthesis and considers the benefits and issues raised by modelling with increased dimensionality. Overall, the challenges raised when building a useful physically modelled voice synthesiser are reviewed and considered with respect to the proposed three dimensional scenario. 1315-1415 5th Dec, 2008 CS103
Christmas Break
Mark Read (CS) Out of the frying pan, into the fire: Using the UML to model autoimmune and regulatory T cell networks: Modelling biological systems using tools such as the unified modelling language (UML) enables the construction of simulations to perform 'in silico' experimentation. The results of this can support and guide wet-lab experimentation, and help advance our understanding of the biological system under study. The CoSMoS project (involving a number of members of the NSC and IS groups) is developing tools, techniques and processes for the modelling, simulation and validation of complex systems. Guided by this project, this talk presents the initial experiences of applying the UML to the modelling of an autoimmune disease (EAE) in mice, and the network of T regulatory cells that prevents it. The UML provides a series of diagram types capable of expressing both static and dynamic relationships between system elements. While it's typical target domain is software systems, we apply it here to the domain of biological systems. This domain shift raises issues with the expressiveness of the UML, which are discussed in this presentation. 1315-1415 23rd Jan, 2009 CS103
Rebecca Newland (Southampton) Using networks to target space debris: Network theory is currently being investigated to identify debris objects that could be removed from orbit around the earth in the future. There are estimated to be over 13,000 pieces of debris larger than 10cm and these create a risk in terms of damaging operational satellites and in terms of colliding with other debris objects which would create even more debris. Therefore, it would be beneficial if the objects that pose the largest risk to the future environment were to be removed. It will be expensive and difficult to remove any objects via 'Active Debris Removal' because of the location of the objects and because new technology needs to be developed for this to happen. The choice of objects to remove needs to be made carefully based on a robust selection process and network theory allows for the development of a quantitative method for this purpose. 1315-1415 6th Feb, 2009 CS103
Tuze Kuyucu (Elec) On the Properties of Artificial Development and Its Use in Evolvable Hardware:The design of a new (yet another!) biologically inspired artificial developmental system will be described. The cell signalling and the system of Gene Regulatory Networks present in biological organisms are modelled, and tailored for tackling real world problems on electronic hardware. Experiments that show the modular behaviour of the developmental system, as well as its ability to solve non-modular circuit problems will be presented. The experimental problems include various n-bit parity circuits, 2-bit multiplier circuits and a scalable generic parity circuit. Further experiments exploring certain properties of the developmental model will be discussed. 1315-1415 20th Feb, 2009 CS103
James Hilder (Elec) Evolving Intrinsic-Variability Tolerant Logic Transistors are shrinking to dimensions where atomic-level structural differences between individual devices on a integrated circuit are resulting in significant threshold-voltage variability, which is not accounted for using traditional simulation techniques. The nano-CMOS project aims to create statistical SPICE-based models for predicted future transistors. This presentation will introduce the project and the work undertaken at York to evolve and optimise standard-cell library logic circuits to be tolerant to intrinsic variability, using a combination of multi-objective GAs and CGP. 1315-1415 6th March, 2009 CS103
Easter Break
David White and Simon Poulding (CS) A Rigorous Evaluation of Crossover and Mutation in Genetic Programming The role of crossover and mutation in Genetic Programming (GP) has been the subject of much debate since the emergence of the field. The question is important from both a theoretical point of view, by supplying evidence in support of models that describe how GP algorithms operate, as well as in practice: by providing guidance to practitioners on setting parameters that enable the most efficient algorithms. In this talk, we discuss our recent rigorous and principled empirical investigation that contributes new evidence to the debate. We compare the efficiency of an algorithm using crossover to one using mutation on six problems common in the GP literature. We will explain how we ensured a fair and objective comparison of these two different GP algorithms, and how the application of robust analysis techniques enabled us to be confident in our conclusions. We also speculate on why some modern, supposedly more efficient, experimental methodologies were not as effective as we had hoped in the context of this comparison. 1315-1415 5th June, 2009 CS103
Susan Stepney (CS) Grammatical Evolution of L-systems: L-systems are parallel generative grammars that can model branching structures. Taking a graphical object and attempting to derive an L-system describing it is a hard problem. Grammatical Evolution (GE) is an evolutionary technique aimed at creating grammars describing the legal structures an object can take. We use GE to evolve L-systems, and investigate the effect of elitism, and the form of the underlying grammar. (This is "another chance to see" the CEC presentation) 1315-1415 19th June, 2009 CS103

Previous seminars can be found here: 2007-08, 2006-07 and 2005-06
Seminar Organiser Jon Timmis