| Date and Location | Speaker | Title and Abstract |
|---|---|---|
| 22 Oct | Dr Bruce Damer CANCELLED |
|
|
5 Nov YCCSA Seminar Room RCH204 |
Paolo Dini Senior Research Fellow, Department of Media and Communications, London School of Economics and Political Science Visiting Research Fellow Biocomputation Laboratory School of Computer Science University of Hertfordshire |
Algebraic Structures of Interaction Computing This talk will outline the conceptual and mathematical challenges in building a theory for a particular form of bio-computing, which we call interaction computing. Interaction computing is inspired by the dynamic order-construction properties of cell metabolism and falls in the middle between a continuous dynamical systems view of the world (e.g. bottom-up and unplanned construction of order generally modelled by the systems of ordinary differential equations used in systems biology) and a behaviour specification view of the world (e.g. top-down software engineering methods, formal methods, etc). This middle ground is inhabited by discrete-mathematical objects such as automata and their languages. In this talk I will show how approaching this problem from an algebraic perspective leads to very interesting insights. The concept of interaction computing rests on architectural hypotheses such as the law of requisite variety from cybernetics, combined with the assumption that it is possible to define a new model of computation based on interacting sequential machines rather than a single Turing machine. Sequential machines are a type of mathematical function that can formalise the specification of behaviour as sequences of outputs based on input histories. Although software engineering methods are quite effective at realising a particular desired behaviour through a suitable structure and set of algorithms, their level of automation could be improved. From this point of view, interaction computing can be understood as aiming to provide a mathematical foundation to the problem of dynamic service composition. Our bio-computing research approaches this question from the point of view of biology and dynamical systems. I will show how by discretising cell metabolic and regulatory pathways we can derive finite-state automata harbouring interesting algebraic structures such as functionally complete permutation groups in their state sets. Whereas automata structure is well-characterised by algebraic automata theory, we are currently evaluating the feasibility of characterising automata behaviour through symbolic dynamics, which can be seen as a formal framework for bridging dynamical systems, automata, regular languages, and coding. During the talk we will build a small automaton together, in order to see first-hand how through its Krohn-Rhodes decomposition the automaton can be understood to define its own abstract number system. I will then show how symbolic dynamics provides a way to see that dynamical systems can also be understood as an even stranger kind of abstract number system. The talk will end with a brainstorm about how any possible formal connections between dynamical systems and automata as abstract number systems might fit within a theoretical framework of interaction computing built around category theory adjunctions that brings all these concepts together and gives a starting point for future research. |
|
19 Nov YCCSA Seminar Room RCH204 |
Andrew Bradley Associate Professor, Biomedical Engineering, School of Information Technology and Electrical Engineering, The University of Queensland |
Group based classification for medical diagnostics Slides (pdf) The purpose of group based classification (GBC) is to determine the class label for a set of test samples, utilising the prior knowledge that the samples belong to same, but unknown class. This can be seen as a simplification of the well studied, but computationally complex, non-sequential compound classification problem. In this talk, I will discuss a number of approaches that extend variants of the nearest neighbour algorithm to implement non-parametric group based classification. The performances of the proposed techniques are evaluated on both synthetic and real-world data sets and compared with techniques that label test samples individually. The results show that, while no one algorithm clearly outperforms all others on all data sets, the proposed group based classification techniques have the potential to outperform the individual based techniques, especially as the (group) size of the test set increases. In addition, it is shown that algorithms that pool information from the whole test set perform better than two-stage approaches that undertake a vote based on the class labels of individual test samples. |
| 3 Dec | No Speaker | |
|
17 Dec YCCSA Seminar Room RCH204 |
Mic Lones |
AlBiNo: Artificial Biochemical Networks The AlBiNo project focuses upon one of the most complex sets of structures found in biological systems: biochemical networks. These structures are fundamental to the development, function and evolution of biological organisms, and are the main factor underlying the complexity seen within higher organisms. Previous attempts to build hardware and software systems motivated by these structures have led to a group of computer architectures which we collectively refer to as artificial biochemical network models. In this talk, I will give a general introduction to artificial biochemical networks, and show how we have used them to carry out state space targeting in numerical dynamical systems and to control the locomotion of simulated robots. |
|
| ||
|
21 Jan YCCSA Seminar Room RCH204 | Adam Nellis |
From meta models to self-modifying code and evolutionary algorithms. Self-modifying computer code is usually seen as a bad thing, but it doesn't have to be. I will try to convince you that self-modifying code has something to do with biology (in particular, evolution). Meta models are a tool I will use to do this. And along the way, I will describe what I've been doing in my PhD: putting self-modifying code into evolutionary algorithms, changing them from "computer programs that evolve solutions to problems" into "computer programs that evolve computer programs that evolve solutions to problems". |
|
4 Feb YCCSA Seminar Room RCH204 |
Martin Drozda Simulation and Modeling group the Leibniz University of Hannover. |
Immune Inspired Approaches to Fault and Intrusion Detection in Ad Hoc
Wireless Networks We review our research efforts related to the computational interpretation of two basic immune mechanisms: co-stimulation and priming. This computational interpretation was done in the scope of misbehavior detection for ad hoc wireless networks. When formulating co-stimulation and priming as a computational approach we abstracted away from the molecular nature of these mechanisms. Instead we concentrated on their logic structure. Needless to say, despite applying such a simplified view, the path to a useful computational interpretation was not straightforward. Our computational interpretation of co-stimulation allows any misbehavior detection system to introduce two desirable properties: energy efficiency and false positives control. Additionally, it introduces the option to exchange energy efficiency for misbehavior detection rate, while keeping the false positives rate unchanged. Priming enables to check whether network operational conditions are within the limits prescribed by priming thresholds. Since priming is based on co-stimulation, it inherits the same basic properties. Bio: Martin Drozda holds a Dr.rer.nat. degree from the Leibniz University of Hannover. In 1998 he was a visiting scholar at BRICS (Basic Research in Computer Science) at Aarhus University, Denmark on a scholarship awarded by the Danish Research Academy. From 1997 to 2002 he was affiliated with the Slovak Academy of Sciences, Institute of Informatics. From 1999 to 2002 he was a research assistant/visiting scientist at Los Alamos National Laboratory, Basic and Applied Simulation Science (CCS-5) group where he participated in several projects covering the areas of high performance computing, theory and application of simulation sciences, decentralized electric power market, and mobile computing. These projects were funded by the US Dept. of Energy, US Dept. of Defense (DARPA), Motorola and other institutions. Since 2003 he is with the Simulation and Modeling group at the Leibniz University of Hannover. Currently, his research focus is within the area of ad hoc and sensor wireless networks. |
|
18 Feb YCCSA Seminar Room RCH204 | Fiona Polack and Elva Robinson |
Parsimos: a research direction for simulation and science. In December, we submitted a cross-disciplinary research proposal (Parsimos) on multi-layer simulation and parallel computing support. The proposal (whether funded or not) represents a research vision for simulation in scientific research. In the seminar, Fiona will outline the motivation and work proposed in Parsimos. Elva will provide an illustration, one of the case studies in the proposal, multi-layer ants. Parsimos is, in one sense, a logical extension to CoSMoS, computer simulation and modelling: we aim to develop principled approaches to simulation, but we want to extend the scope and scale of what is possible. We are interested in integrating existing models with new models, integrating mathematical and individual (agent style) models, simulating across more than two adjacent scales or levels, and being able to substitute models within a simulation to support extraction and exploration of hypotheses. There is lots of potential for cross-disciplinary research - and, if this proposal is not funded, we will be actively seeking more collaborative and funding opportunities very soon! |
|
4 Mar YCCSA Seminar Room RCH204 | Martin Trefzer |
PAnDA - Programmable Analogue and Digital Array
Not Everything is Black & White The PAnDA project focuses upon one of the currently greatest challenges in nano-scale electronic design: reliably creating functional circuit designs when device sizes approach atomistic scales. Device physics are harder to model in deep sub-micron processes as intrinsic stochastic variability affects circuit design. The PAnDA project aims to develop understanding of the effects of stochastic variability and to propose novel design methodologies to overcome these intrinsic variations. PAnDA will involve the design and fabrication of a novel reconfigurable variability tolerant architecture, which allows variability aware design and rapid prototyping by exploiting the configuration options of the architecture. These are vital steps towards the next generation of FPGA architectures. |
|
18 Mar YCCSA Seminar Room RCH204 | Jon Timmis |
Toys, Toys and more Toys (I mean robots for serious academic work) In May a new robotics laboratory will be opened. During this talk, I will first discuss some of the immune-inspired robotics work that we have been developing over the past few years that will, we hope, in part populate our new laboratory. Work ranging from fault tolerant swarm systems based on the formation of granulomas, to general immune-inspired architectures for fault tolerance in swarm robotic systems, to detecting improvised explosive devices. We will talk about future planned work in the area, including how to make use of T-regulatory networks involved in recovery from EAE (a mouse proxy for MS) in the co-ordination of underwater swarm robotic systems. I will then open the floor for discussions on ideas for possible work that people might be interested in doing in the lab in the future. To help support this, I will provide a short overview of the robots that we currently have that people might want to make use of and have a discussion and see where it leads us. |
|
| ||
|
6 May YCCSA Seminar Room RCH204 |
Inman Harvey Work done in collaboration with Nick Tomko |
Artificial Evolution meets Metagenomics: the Binomic GA for Evolving Symbiotic Systems Evolutionary Algorithms (such as Genetic Algorithms) are typically based on a conventional picture of evolution with vertical genetic transfer, from one generation to the next, and fitness being calculated for each individual. However very recent Metagenomic studies have painted a rather different picture of the microbial evolution that has comprised the first 2 billion years of evolution on this planet, together with the majority of evolution even today. For instance, a human being is a walking ecosystem that has coevolved with the microbial cells within; only 1% of the cells within the typical human-microbial ecosystem are actually human, and they are all coexisting in symbiosis. So this new picture of evolution is one of horizontal gene transfer -- genes are swapped between microbes without the need for successive generations; and evaluation of fitness is at the symbiotic or community level, based on the viability of an ecosystem. Though one or other of these aspects have featured in some evolutionary computation before, here in the Binomic GA we incorporate both together for the first time. |
|
20 May YCCSA Seminar Room RCH204 1315 - 1415 |
Steve Smith |
Differential Diagnosis of Neurodegenerative Diseases using Evolutionary Algorithms Alzheimer’s disease, Lewy body dementia and Parkinson’s disease with and without dementia are four of the most common neurodegenerative brain disorders, affecting in excess of 700,000 people in the UK alone. All four disorders can exhibit similar symptoms and consequently, a reliable diagnosis can be difficult to obtain. No reliable clinical test is available and yet accurate, early, diagnosis is essential to ensure the appropriate treatment or therapy is administered to patients in a timely and cost efficient manner. Previous laboratory-based experiments have demonstrated some success in differentiating between these conditions, but are unsuitable for reliable diagnosis in a clinical environment. This presentation will describe a clinical trial scheduled to begin in July at Leeds General Hospital that will employ commercially available sensor technology in conjunction with evolutionary algorithms to differentiate between these conditions in a reliable and safe way, that is compatible with existing clinical practice. |
|
3 Jun YCCSA Seminar Room RCH204 | S. Barry Cooper |
The 'Mathematician's Bias', and the Return to Embodied Computation Slides (pdf) In the ACM Ubiquity Symposium on "What Is Computation", Dennis Frailey (see http://ubiquity.acm.org/article.cfm?id=1891341 ) refers to the "mathematician's bias" (being an emphasis on the computability of functions rather than processes) - which "limits our thinking and prevents us from fully appreciating the power of computation ... Today, I believe we are breaking out of the era where only algorithmic processes are included in the term computation." In this talk we explore the theme of the mathematician's bias, and some available routes to renewing the relevance of classical computability. |
|
17 Jun YCCSA Seminar Room RCH204 |
Ed Clark |
Recurring Floating Point Binary Tent maps are a type of discrete time dynamical system. They have interesting mathematical properties including chaos. We introduce tent maps and use them to demonstrate an interesting problem with calculations using floating point numbers. We demonstrate that certain starting values that provably should go to periodic solutions quickly lose numerical resolution in a floating point binary representation. We go on to discover that it is not simply a matter of insufficient accuracy (mantissa length) - any finite accuracy is insufficient. Faced with a proper understanding of such a problem, the solution is simple: you need infinite accuracy in your floating point binary numbers. We explain how this can be achieved for recurring binary numbers. We introduce a representation for numbers in computers: recurring floating-point binary. We describe the properties of the system and how they are a solution to the problems encountered in computation of tent maps. We then compare and contrast recurring floats with the standard IEEE 754 floating point binary and "a/b" (where "a" and "b" are integers). |
| 1 Jul | ||