SEMINARIESCHEMA FÖR MATEMATISK STATISTIK 

Fredag    29/9  15.15         Isaac Meilijson,
          1996                School of Mathematical Sciences,
                              Raymond and Beverly Sackler Faculty of
                              Exact Sciences, Tel Aviv University

                              A Hidden Markov Model for neuronal firing
                              while concentrating on a task
Abstract:

Hidden Markov models were introduced in the late sixties by Baum, Petri 
and others as models for imperfectly observed Markov chains. These models 
were generalised in the late eighties to Influence Diagrams by Lauritzen, 
Spiegelhalter and others, as models for diagnostic Bayesian updatings of 
distributions of random variables that may signify in medical applications
diseases and results of laboratory tests, but apply more generally to any 
other setup involving the probabilistic updating of many random factors 
that are selectively observed. These models belong to the general
statistical framework of Incomplete Data models, as formali\-zed in the 
seventies by Dempster, Laird and Rubin, and its techniques are versions,
applications or modifications of the EM algorithm. 

The talk will be devoted to an illustration of the hidden Markov model to 
an attempt at statistically detecting ``mental states'', or sequences of 
hidden quasi stationary states, 
in the cortical activity of a monkey that concentrates on remembering the 
correct action (out of four possibilities) to pursue later, when given a 
command. 
This work is based on measurements performed by a team of biologists at 
the Computational Neurobiology Laboratory at the Hebrew University of 
Jerusalem, under the direction of Moshe Abeles, with whom the speaker 
co-supervised Eyal Seidemann in the performance of this analysis.    




Lokal: Rum 227 i Mattehuset.

Björn Holmquist 046-2228546