Computer Go

Alfredo Braunstein

Go is an ancient Chinese game that originated some 4000 years ago and has still great popularity nowadays. Computer Go on the other hand has made little progress in these 4000 years: best go programs are rated like middle-to-weak amateur human players. We will discuss one recent approach to computer go [¹], based on a mixture of two relatively well known strategies: the UCT algorithm and Monte Carlo, which happens to be the most successful one to date.

[1] Modification of UCT with Patterns in Monte-Carlo Go, S. Gerlly, Y. Wang, R. Munos and O.Teytaud, INRIA Technical Report, 2006.

A reliability index for clades, based on taxonomical congruence

Blaise Li

I will present part of my PhD work in phylogeny. Phylogeneticists use comparative data to reconstruct the "genealogy" of taxa (usually, species or genuses): a "phylogeny". The data can be morphological characters, DNA sequences, etc. A practical problem is that on large groups, different datasets tend to produce trees that are not the same. By comparing trees obtained from different independent datasets, one can get an idea of which clades (groups of taxa) are reliable and which are not. The more a group is repeated, the more it is reliable. I tried to formalize this principle into a "repetition index" that can be attributed to clades. One essential difficulty is that datasets often contain missing data; therefore, trees are not built on the same set of species.

Efficient approximation of Gaussian Mixture product for nonparametric belief propagation algorithm

Limin Fu

Gaussian mixtures have been used in many mathematical and computational models to approximate arbitrary distributions. Nonparametric Belief Propagation (NBP) algorithms use Gaussian mixtures to approximated the message and belief distributions in cases where discretized representations may become unfeasible. The key issue in Gaussian mixture based NBP is the computation of products of Gaussian mixtures. In this talk, I will introduce the approach I developed to approximate Gaussian mixture product efficiently based on a Gaussian mixture reduction technique. This approach was integrated into a NBP algorithm with a simple model for the stereo matching problem.

A review of the locust olfactory bulb

Carlo Baldassi

A review of the locust olfactory bulb, based on the following papers: 1) Gilles Laurent, Katrina MacLeod, Mark Stopfer, et al. Spatiotemporal Structure of Olfactory Inputs to the Mushroom Bodies Learn. Mem. 1998. 2) Stijn Cassenaer and Gilles Laurent Hebbian STDP in mushroom bodies facilitates the synchronous flow of olfactory information in locusts Nature 2007. 3) Laurent Moreaux and Gilles Laurent Estimating firing rates from calcium signals in locust projection neurons in vivo frontiers in neural circuits 2007.

How to infer gene networks from expression profiles

Valentina Lanza

Inferring, or ‘reverse-engineering’, gene networks can be defined as the process of identifying gene interactions from experimental data through computational analysis. Gene expression data from microarrays are typically used for this purpose. Here we compared different reverse-engineering algorithms for which ready-to-use software was available and that had been tested on experimental data sets. We show that reverse-engineering algorithms are indeed able to correctly infer regulatory interactions among genes, at least when one performs perturbation experiments complying with the algorithm requirements. These algorithms are superior to classic clustering algorithms for the purpose of finding regulatory interactions among genes, and, although further improvements are needed, have reached a discreet performance for being practically useful.

[1] M. Bansal1, V. Belcastro, A. Ambesi-Impiombato and D. di Bernardo, Molecular Systems Biology, 3(10.1038/msb4100120), 2007.

Add to calendar
Syndicate content