Graduate Talks MT 2008
The format is 15 minutes for the talk, followed by 5 minute for discussion
Schedule of Talks: Lecture Room (1 South Parks Road)
2:00 Bo Chen
Title: Partly exchangeable paint-box construction
Abstract: Kingman's paint-box construction has been widely used in phylogenic models to describe the splitting behaviour. But it only applies for the model with exchangeable leafs such as Beta model and model related to stable trees.
We extended the Paint-box construction to partly exchangeable trees which covers the argument of Alpha model and Alpha-Gamma model.
2:20 Joanna Davies
Title: A conceptual framework common to studies of complex phenotypes.
Abstract: It comprises of three components; data, concepts and analyses.
Traditionally, studies were founded on the collection of one or two data types, but with the recent emergence and deployment of affordable high-throughput biological technologies, an increasing number of studies are reporting integrative approaches. I will evaluate the contributions of integrative studies to biological understanding and conclude with discussion about integrative functional studies as a necessary follow-up to global discovery driven approaches.
2:40 Aziz Mithani
Title: A stochastic model for network evolution.
Abstract: The evolution of metabolic networks is characterized by gain and loss of reactions (or enzymes) connecting two or more metabolites and can be described as a discrete space continuous time Markov process. We introduce a neighbour-dependent model for the evolution of metabolic networks where the rates with which reactions are added or removed depend on the fraction of neighbouring reactions present or absent from the network. In this model, metabolic networks are represented as directed hypergraphs, where an edge
(reaction) may connect any number of vertices (metabolites). This provides an intuitive approach to study evolution since loss or gain of reactions can be regarded as loss or gain of hyperedges. An MCMC algorithm for sampling paths between two networks is also presented.
3:00 Elizabeth Ford
Title: The distribution of the degree of a randomly chosen vertex in a Barabasi-Albert random graph
Tea 3:20 - 3:40
3:40 Shahzia Anjum
Title: A Boosting Approach to Network Inference
Abstract: Our approach termed as BoostiGraph is a simple and computationally cheap algorithm which can be used to learn the structure of high dimensional network structures. Our method complements the Lasso based approaches like glasso. We demonstrate via simulations that our approach is capable of performing as well as glasso and GeneNet (another network inference approach) whilst being extremely fast. l 4:00 Rhodri Saunders Title: Bias In Protein Structure Prediction The accurate prediction of both secondary and tertiary protein structure is an ongoing problem. With a view to ultimately improving protein structure prediction we analyse structure predictions for potential algorithm bias. We find that structure prediction is significantly more accurate at the amino(N)-terminus compared to the carboxy(C)-terminus.
4:20 Ravi Kalia
Title: Ordinal classification with reproducing kernel Hilbert spaces.
Abstract: A brief introduction to a multivariate SVM approach to classifying patterns. SVMs have been used to classify patterns in the binary case with estimation done via the solution of a Quadratic Programming problem. We consider a mechanism to simultaneously estimate ordinal classification rules from the same reproducing kernel Hilbert space.
Printer Friendly Version 