Seminar Programme
Tea and coffee are served after the seminars in Room 115, McCrea Building. All are welcome.
Other seminar series of interest
(All seminars will be held on Tuesdays from 1-2, followed by tea in the Journals room, unless otherwise stated).
Spring 2010 Seminars
- 23 February 2010
Dave Cohen, RHUL: How hard is optimisation?
1pm, C325
The Constraint Satisfaction Problem (CSP) is concerned with the feasibility of satisfying a collection of constraints. The CSP paradigm has proven to be useful in many practical applications. In a CSP instance a set of variables must be assigned values from some domain. The values allowed for certain (ordered) subsets of the variables are restricted by constraint relations. The general CSP is NP-hard. However there has been considerable success in identifying tractable fragments of the CSP. In particular, the tractability of the CSP fragments obtained by limiting the allowed set of constraint relations is characterised by an algebraic property: the clone of polymorphisms. The extension of this framework, called the Valued CSP (VCSP), which allows optimisation, is now also being investigated. In this extended framework, for some ordered subsets of the variables, each possible tuple is given a (possibly infinite) desirability weighting, or cost. The goal is to find the most desirable (or least cost) total assignment. We have replaced constraint relations, which define feasible tuples, with cost functions, which give desirability values to tuples. In this talk we will show that, when the costs are rational numbers or infinite, the complexity of the VCSP fragments obtained by limiting the cost functions is determined by certain algebraic properties: feasibility polymorphisms and fractional polymorphisms. As an immediate application of these results, we show that the existence of a non-trivial fractional polymorphism is a necessary condition for tractability (assuming P != NP).
- 2 March 2010
Alexey Chernov, RHUL: Multiobjective prediction with expert advice?
1pm, C325
Prediction with Expert Advice is a popular model for sequence prediction. At each step, Experts give their predictions, then Learner gives his prediction that may depend on Experts' opinions, and then the outcome is revealed. The difference between predictions and outcomes is evaluated by some loss function. Learner's goal is to keep his accumulated loss small compared to the accumulated losses of each Expert. We consider a new framework, where the quality of predictions is evaluated not by a single loss function but by a (finite) set of them. Accordingly, Learner has a set of losses, and his new goal is to keep all his losses small. We construct a strategy for Learner that achieves this goal if the loss functions are mixable and proper. This is a joint work with Volodya Vovk.
- 16 March 2010
Richard Morris, John Innes Centre: Biological information processing: Unravelling the control logic of the floral transition
1pm, WIN003
I will present the computational approaches (network modeling, parameter space sampling, machine learning) we developed to model flowering time. The floral transition involves the switch from vegetative to reproductive growth and is an important adaptive trait in plants. After the floral transition, groups of cells on the flanks of the shoot apex (primordia) are specified to form flowers instead of leaves. Arabidopsis thaliana integrates environmental and endogenous signals to determine the correct timing of flowering. Once initiated, the floral transition is rapid and irreversible on the flanks of the shoot, while nearby cells in the centre of the shoot are maintained in an undifferentiated state to support future meristematic growth. The regulatory logic that enables simultaneous irreversible switching in floral primordia and the maintenance of undifferentiated cells in the shoot is not understood. We have developed a mathematical model that captures important features of the floral transition with which we can quantitatively account for all known mutants. The model makes valuable predictions for novel genotypes that have recently been validated and explains a known phenotype that was described many years ago but was poorly understand.
Summer seminars 2010
- May 18, 2010 Marek Sergot, Imperial College, London
- May 25, 2010 Qiang Shen, Aberystwyth University
Recent seminars
| Date | Friday 11th September 2009 |
| Speaker | Dr. Jüri Lember - University of Tartu |
| Location | McCrea 219 |
| Title | On recent developments in HMM-based segmentation |
| Abstract | Hidden Markov models (HMM) are commonly used in many areas, including
signal processing, speech recognition and computational molecular
biology. Roughly, an HMM consists of an unobserved hidden Markov
chain
and its "noisy" observed counterpart. Often, the main object of
interest is the true realization of the hidden Markov chain from
observed data. Estimation of the hidden realization is also known as
segmentation. For example, speech signal can be segmented into
sentences, words, phonems etc; DNA sequences can be segmented into
coding and non-coding regions, or into CpG islands and background. A
widely used estimate of the underlying hidden path is
maximum-likelihood, or maximum a posteriori (MAP), sequence of
states,
commonly known as Viterbi alignment. A well-known alternative is
pointwise, i.e. symbol-by-symbol, maximum-likelihood, or
PMAP-alignment. This latter alignment is optimal in the sense of
minimizing the expected number of incorrectly classified components
of
the hidden sequence. However, depending on the task, the best
(optimal) alignment might be yet different from those two.
When the HMM extends to infinity, then (under certain mild
conditions)
many alignments (including Viterbi and PMAP alignments) converge,
generally to distinct limiting alignments. Viewed as proper random
processes derived from the original hidden Markov process, the
infinite limiting alignments can be used to study various asymptotic
properties of their respective segmentations.
We introduce a somehow generalized view of segmentation, and we show how the asymptotic analysis can be used for measuring and improving the goodness of segmentation. This is a joint work with Alexey Koloydenko. |
| Date | 15th September 2009, 13:00 |
| Speaker | Dr. Robin Adams - Department of Computer Science, Royal Holloway |
| Location | McCrea 219 |
| Title | Type Theory and Proof Checkers |
| Abstract | A type theory is a formal language which deals with statements M:A. Remarkably, the same formal language can be read in three different ways: * M is an object of type A * M is a program that satisfies the specification A * M is a proof of the proposition A This minor miracle - the fact that the same formal rules govern all three interpretations - goes by the name of the Curry-Howard Isomorphism, and is a very useful fact that is still not fully understood. In this talk, I shall give a non-technical survey the history of type theories, and explain how they are useful for building proof checkers - systems that decide whether a proof of a mathematical theorem is correct. I shall describe my own research with logic-enriched type theories, which can be seen as hybrid systems between type theory and predicate logic. |
| Date | 29th September 2009 |
| Speaker | Prof. Dr. Vladimir V. Anisimov Director, Research Statistics Unit GlaxoSmithKline |
| Location | TBA |
| Title | Statistical modelling in clinical trials (patient recruitment and drug supply) |
| Abstract | A large clinical trial for testing pharmaceutical drug usually
involves hundreds of patients to be recruited by many clinical
centres. Patients recruited into the trial are randomized to prescribed
medical treatments and at some stages, their responses are
statistically analysed for taking a decision on the effectiveness of
the drug. This process requires using various statistical techniques at
different stages of the trial.
In the first part of the talk, the basic statistical problems that appear in drug development process in pharmaceutical industry are briefly described. The second part is dealing with the patient recruitment stage. This stage is very costly and affected by various uncertainties in input information, variation in clinical centres and stochastic fluctuations of the recruitment over time. A large proportion of trials fail to complete recruitment before deadline. A novel statistical methodology for modelling and predicting patient recruitment and drug supply in multicentre clinical trials is proposed. Patient's flows in different centres are viewed as delayed Poisson processes with random unknown rates. The problems of modelling, estimating and predicting with credibility bounds the number of patients over time and the recruitment time are considered. The developed technique also allows re-assessing the number of clinical centres required to complete recruitment in time with a given confidence (adaptive adjustment), predicting study/centre performance, and evaluating the amount of drug supply which is needed to cover patient demand with a given risk of undersupply. The technique has been validated for many real trials and is on the way of implementation. Illustrative examples are considered. References: V. Anisimov, V. Fedorov, Modeling, prediction and adaptive adjustment of recruitment in multicentre trials, Statistics in Medicine, 26, 27, 2007, pp. 4958-4975. V. Anisimov, Recruitment modeling and predicting in clinical trials, Pharmaceutical Outsourcing, March/April 2009, pp 1-10. |
| Please note that the following seminar does not happen at the usual time or place | |
| Date | Friday 4th December 2009, at 4pm |
| Speaker | Dr.G.Martynov, The Institute for information transmission problems of the Russian Academy of Sciences (Kharkevich Institute). |
| Location | McCrea 219 |
| Title | Karunen-Loeve decompositions for weighted Wiener process and for Brown bridge with applications to the goodness-of fit tests |
| Abstract | We will consider the Wiener process and Brown bridge with the weight fuction t^\alpha. The decomposition of weighted processes was fulfilled with usig of Bessel functions. It was found the eigen values and eigen fuctions for the Fredholm operator with the corresponding covariance functions as kernel. These results have the applications to the Kramer- von Mises goodness of fit test. The results were found in cooperation with P.Deheuvels (University Paris VI). |
Other seminar series of interest :
Information Security Group
Department of Mathematics
Department of Physics
School of Biological Sciences
Gatsby Institute, UCL
Imperial College Bioinformatics Seminars
For previously held seminars, please see our Events Archives.
Tel/Fax : +44 (0)1784 443421 /439786