SEMINARIOS DE INVESTIGACIÓN EN INGENIERÍA INFORMÁTICA Y DE TELECOMUNICACIÓN 2007-2008

 

Actividad de Formación Continua  del Programa Oficial de Posgrado en Ingeniería Informática y de Telecomunicación

Escuela Politécnica Superior, Universidad Autónoma de Madrid

                       

jueves, 4 de diciembre de 2008, 12:30

Salón de Grados, Escuela Politécnica Superior, Universidad Autónoma de Madrid

Bayesian Machine Learning Methods for Computational Neuroscience

Terran Lane

University of New Mexico, Department of Computer Science
     

Abstract

The field of neuroimaging has developed an amazing array of equipment for noninvasively sensing the brain's activity: magnetic resonance imaging, magnetoencephalography, electroencephalograpy, near-infrared imaging...  These technologies are providing us unprecedented views of the behavior of the living brain.  However, they also produce astounding volumes of noisy, structured, complicated data that is only indirectly tied to the underlying signal of interest (neural activity).  The task of converting such complex data sources into neuroscience knowledge is daunting.

Much of my current research focuses on developing data analysis and inference algorithms for extracting neuroscientifically significant patterns from complex neuroimaging data sources.  In this talk, I will give a brief overview of the functional magnetic resonance imaging (fMRI) and magnetoencephalography (MEG) technologies, outlining some of the challenges in their analyses.  I will describe the Bayesian inferential approach that my research group has been taking to the analysis of these data sources and will outline some of the statistical and algorithmic challenges that arise in these analyses. Specifically, I will describe our methods for identification of brain activity networks from indirect data and for sensor fusion of multimodal neuroimaging data sources.

PDF presentation

Terran Lane

Terran Lane is assistant professor of computer science at the University of New Mexico. His personal research interests include machine learning, behavioral modeling and learning to act/behave (i.e., reinforcement learning), scalability, representation, and the tradeoff between stochastic and deterministic modeling. All of these represent different facets of his overall interest in scaling learning methods to large, complex spaces and using them to learn to perform lengthy, complicated tasks and to generalize over behaviors. While he attempts to understand the core learning issues involved, he often situates his work in domain studies in practical (well, ok, semi-practical anyway) problems. Doing so both elucidates important issues and problems for the learning community and provides useful techniques to other disciplines. More information: http://www.cs.unm.edu/~terran/