Randy Moths and Hungry Snails on the Edge: Pushing the Critical Brain Hypothesis Beyond the Cerebral Cortex
Location: D'Arcy Thompson Room, School of Computing Sciences, UEA
Date: 14.00 10 Dec 2010
Speaker: Chris Buckley
Organiser: Dr. Hywel Williams
Institution: University of Sussex
Abstract: The hypothesis that biological systems exhibit dynamics critically poised at the boundary between order and chaos was popularised by Stuart Kauffman almost 20 years ago now. Recently, neuroscientists have begun to take this idea seriously and there have been several experiments that have suggested that the statistical dynamics of the mammalian brain exhibit the hallmarks of criticality. However, all experimental work thus far has been done in the cerebral cortex. Yet if we are to take Kauffman's original ideas seriously it should be possible to examine the idea of critical dynamics in other biological systems.
In this talk I retain a focus on "neural" criticality but discuss the possibility that it could play a central role in the dynamics of invertebrate sensory and motor systems. To make progress on this issue I present a different theoretical framework for biological criticality inspired by the field of synergetics and based on the idea of bifurcation. I will then apply these ideas to a biologically detailed model of the moth pheromone system and show how neural criticality could explain how sexually motivated male moths are able to locate females from more than a mile away. I will then briefly outline a new collaboration with biologists working on the pond snail Lymnaea and discuss the idea that the hungry snail's feeding system is poised at a critical point. I finish by discussing the prospect of establishing the ubiquity of critical dynamics in neural systems and the possible implications this could have for how we model and understand their behaviour.

Topological Properties of Gene Regulatory Networks with Qualitatively Different Gene Expression Dynamics in Spatially Organised Systems
Location: D'Arcy Thompson Room, School of Computing Sciences, UEA
Date: 14.00 26 Nov 2010
Speaker: Dr Costas Bouyioukos
Organiser: Dr. Hywel Williams
Institution: Sainsbury Laboratory
Abstract: In biological systems gene expression within a cell is determined by a network of regulatory interactions among genes mediated by gene products.
In spatially organised systems consisting of multiple cells, gene expression within a cell is also affected by gene activity taking place in neighbouring cells, through exchange of gene products. The interplay between spatial organisation between cells and the gene regulatory network (GRN) within each cell may qualitatively alter gene expression dynamics and has an impact on, spatially extended, essential biological processes such as cell differentiation, pattern formation and morphogenesis.
This talk will present a computational framework to model GRNs on spatially organised systems and investigate the effects of GRN topology on their pattern formation potential. GRNs are scored by comparing the level of gene expression heterogeneity on a lattice to the levels generated on a well stirred reactor as a null model. That score is used to quantify the propensity of GRNs to form patterns on spatially organised systems.
A set of network topological properties, such as density and network centralities, that are used to characterised biological networks are correlated with the pattern formation capacity. Associations between the capacity of GRNs to form spatial patterns and their topological properties will be discussed.
Robustness of the pattern formation capacity of GRNs will be investigated against random initial conditions variation, as well as random and directed deletion of network elements, such as nodes and edges.

Weighted scores method for regression models with dependent data
Location: D'Arcy Thompson Room, School fo Computing Sciences, UEA
Date: 12.00 22 Oct 2010
Speaker: Dr Aristidis Nikoloulopoulos
Organiser: Dr. Hywel Williams
Institution: School of Computing Sciences
Abstract: There are copula-based statistical models in the literature for regression with dependent data such as clustered and longitudinal overdispersed counts, for which parameter estimation and inference is straightforward. For situations where the main interest is in the regression and other univariate parameters and not the dependence, we propose a weighted scores method, which is based on weighting score functions of the univariate margins. The weight matrices are obtained initially fitting to the data a discretized multivariate normal distribution, which admits a wide range of dependence. We present the application of our general methodology to negative binomial regression models. Asymptotic and small sample efficiency calculations show that our method is robust and nearly as efficient as the maximum likelihood for fully specified copula models. An illustrative example is given to show the use of our weighted scores method to analyze utilization of health care based on characteristics of the families.

Automatic characterization of materials appearance through texture and colour analysis
Location: D'Arcy Thompson room, School of Computing Sciences, UEA
Date: 14.00 15 Oct 2010
Speaker: Francesco Bianconi
Organiser: Dr. Hywel Williams
Institution: University of Perugia (Visiting Fellow, School of Computing Sciences, UEA)
Abstract: Automatic evaluation of the visual appearance of materials plays an important role in many applications, such as surface grading, defect detection, content-based image retrieval, etc. The speaker, who is currently visiting fellow in the School of Computing Sciences at the UEA, will be discussing some theoretical and practical issues related to these topics.

Flat Split System - An Introduction and Application
Location: D' Arcy Thompson Room, School of Computing Sciences, UEA
Date: 14.00 29 Sep 2010
Organiser: Dr. Hywel Williams
Institution: School of Computing Sciences, UEA
Binh Nguyen will present his PhD research on phylogenetics, which was supervised by Prof Vincent Moulton.
Abstract: Split systems are a popular tool in phylogenetics: They are the underlying object for phylogenetic trees, phylogenetic (split) networks and thus, problems concerning these, i.e. reconstruction, optimisation.
Several classes of split systems, namely compatible, circular, and affine split systems, have been being studied in both theory and practice. We present new results about a new class of split systems called flat split system (FSS). We will define the FSSs and show how a planar split network can be constructed to represent a FSS. We then define operations on the FSSs and use them to search for a weighted FSS that represents a given distance matrix. The weighted FSSs will be compared to corresponding circular split system produced by the popular algorithm NeighborNet.

Speech/Vision Talks on Friday, 24th Sept. 2010
Location: D'Arcy Thompson Room, School of Computing Sciences, UEA
Date: 11.00am 24 Sep 2010
Organiser: Prof. Stephen Cox
Institution: UEA
Showcase event highlighting current research in the Graphics, Vision and Speech Laboratory in the School of Computing Sciences

The Wisdom of Solomon- a brief History of Variance Stabilization and its connection to Kullback-Liebler Divergence
Location: D'Arcy Thompson Room, UEA
Date: 14.00 1 Jul 2010
Speaker: Prof Robert Staudte
Organiser: Dr. Hywel Williams
Institution: La Trobe University, Australia
Abstract: In the beginning was R.A. Fisher, who created variance stabilization and saw that it was good. Many statistical descendants have found it to be a powerful tool in applications. In a recent book Kulinskaya, Morgenthaler and Staudte (Wiley, 2008) formalized this powerful tool to define statistical evidence on a scale that allowed for simple calibration and interpretation, as well as combination of evidence from different studies in a meta-analysis. It turns out that the Key Inferential Function of their book that arises in variance stabilization has a remarkably strong link to the Kullback-Liebler Divergence. This result leads to some surprising global approximations and new test statistics.

Extracting structure from galaxy images
Location: D'Arcy Thompson Room, School of Computing Sciences, UEA
Date: 14.00 21 May 2010
Speaker: Dr. Wayne Hayes
Organiser: Dr. Hywel Williams
Institution: Imperial College London
Abstract: The amount of data from sky images is large and growing.
About 1 million galaxies can be discerned in the Sloan Digital Sky Survey (SDSS); the Large Synoptic Survey Telescope (LSST) is being built and will scan the entire sky repeatedly, providing images of millions of galaxies and terabytes of data every night; and the Joint Dark Energy Mission (JDEM) is a proposed orbiting satellite that will repeatedly map the entire sky from orbit, providing images of perhaps billions of galaxies. Unfortunately, given an image of a spiral galaxy, there does not exist an automated vision algorithm that can even tell us which direction the spiral arms wind, much less count them or provide any other quantitative information about them. To wit, the largest current galaxy classification project is the Galaxy Zoo, in which thousands of human volunteers classify images by eye over the web.
Although valuable, such human classifications will provide only limited objective quantitative measurements, and will soon be overwhelmed with more data than humans can handle. However, such information would prove an invaluable source for astronomers and cosmologists to test current theories of galaxy formation and cosmic evolution (which can now be simulated with high accuracy on large computers, producing copious predictions that cannot be tested due to a lack of objective, quantitative observational data). In this talk, I will report on preliminary results from dynamical grammars and other machine learning and vision techniques to "parse" images of galaxies, starting us on the road towards producing quantitative data that will be useful for astronomers to test theories of spiral galaxy formation, evolution, and merging.

A Measurement Error Model for Heterogeneous Capture Probabilities
Location: D'Arcy Thompson Room, School of Computing Sciences, UEA
Date: 12.00 30 Apr 2010
Speaker: Prof Richard Huggins
Organiser: Dr. Hywel Williams
Institution: University of Melbourne
Abstract: A Measurement Error Model for Heterogeneous Capture Probabilities in Mark-Recapture Experiments.
Mark-recapture experiments are used to estimate the size of animal populations.Logistic models for capture probabilities that depend on covariates are effective if the covariates can be measured exactly. If there is measurement error so that a surrogate for the covariate is observed rather than the covariate itself, simple adjustments may be made if the parameters of joint distribution of the covariate and the surrogate are known. Here we consider the case when a surrogate is observed whenever an individual is captured and the parameters must also be estimated from the data. A regression calibration approach is developed and it is illustrated on a data set where the surrogate is an individual bird's wing length.

An Introduction to the UK and European Games Industries - AMD Graphics Products Group
Location: D'Arcy Thompson Room, School of Computing Sciences, UEA
Date: 14.00 19 Mar 2010
Speaker: Kevin Strange, Developer Relations Account Manager
Organiser: Dr. Hywel Williams

How to merge normal mixture components for cluster analysis
Location: D'Arcy Thompson room, University of East Anglia
Date: 14.00 26 Feb 2010
Speaker: Christian Hennig
Organiser: Dr. Hywel Williams
Institution: University College London
Abstract: Normal mixture models are often used for cluster analysis. Usually, every component of the mixture is interpreted as a cluster. This, however, is often not appropriate. A mixture of two normal components can be unimodal and quite homogeneous. Particularly, mixtures of several normals can be needed to approximate homogeneous non-normal distributions.
Normal mixture models are often used for cluster analysis. Usually, every component of the mixture is interpreted as a cluster. This, however, is often not appropriate. A mixture of two normal components can be unimodal and quite homogeneous. Particularly, mixtures of several normals can be needed to approximate homogeneous non-normal distributions.
Even if there are non-normal subpopulations in the data, the normal mixture model is still a good tool for clustering because of its flexibility. This presentation is about methods to decide whether, after having fitted a normal mixture, several mixture components should be merged in order to be interpreted as a single cluster.
Note that this cannot be formulated as a statistical estimation problem, because the likelihood and the general fitting quality of the model does not depend on whether single mixture components or sets of mixture components are interpreted as clusters. So any method depends on a specification of what the user wants to regard as a "cluster". There are at least two different cluster concepts, namely identifying clusters with modes (and therefore merging unimodal mixtures) and identifying clusters with clear patterns in the data (which for example means that scale mixtures, though unimodal, should not necessarily be merged). Furthermore, it has to be specified how strong a separation is required between different clusters.
The methods proposed and compared in this presentation are all hierarchical.
From an estimated mixture, pairs of components (and later pairs of already merged mixtures) are merged until members of a pair are separated enough in order to be interpreted as different clusters. This can be measured in many different ways, depending on the underlying cluster concept.
Apart from the discussed methodology, some implications about how to think about cluster analysis problems in general will be discussed.

An Introduction to Physics-based Computer Vision
Location: D'Arcy Thompson Room, School of Computing Sciences, UEA
Date: 14.00-15.00 19 Feb 2010
Speaker: Dr. Robby Tantowi Tan
Organiser: Dr. Hywel Williams
Institution: Utrecht University, The Netherlands
Abstract: The world contains abundant visual information. The main task of computer vision is to acquire, extract and interpret that information into meaningful representations.
The world contains abundant visual information. The main task of computer vision is to acquire, extract and interpret that information into meaningful representations. Over decades of development, computer vision has become a field with multiple aspects and approaches emerging from diverse disciplines and applications. One of the important approaches is physics-based computer vision (or physics-based vision, for short) which in principal utilizes physical models of the world to extract visual information. This approach has become one of the core topics in computer vision, since visual information is in fact spatial collections of light rays that are emitted, transmitted, scattered, absorbed, and reflected according to physical laws. While it has significantly contributed to the development of computer vision, physics-based vision has also played crucial roles in computer graphics and virtual reality, since to arrive at realistic image rendering, one should employ physical models of the world and their parameters.

Computational Models for Colour Naming
Location: D'Arcy Thompson Room, School of Computing Sciences, UEA
Date: 12.00-13.00 29 Jan 2010
Speaker: Robert Benavente
Organiser: Dr. Hywel Williams
Abstract: Colour is an important visual cue widely used in computer vision. Most of the methods developed so far aim to extract low-level colour features from images and such information do not have a direct link to the high-level semantics that humans use. The lack of this link, known as the semantic gap, is even more significant for some applications such as image retrieval, in which users require systems supporting natural-language queries. Hence, the goal of our research is to reduce the semantic gap in the task of giving names to the colours in an image.
The problem is framed on the fuzzy set theory and each of the 11 basic colour categories (white, black, red, green, yellow, blue, brown, purple, pink, orange, and grey) is characterized by a parametric membership function. After fitting the parameters of the model, it is possible to compute the membership of any pixel to the 11 colour categories considered.
Our first model, defined in calibrated conditions, obtained good results on an image annotation task. However, several problems, such as changes in the illumination, presence of shadows and other contextual effects, challenge the success of the model on other real applications. A first step towards overcoming these problems has been done by redefining the model in a colour appearance model which takes into account the viewing conditions and the surround effects on any pixel of an image.