thesis available

[Yoram Singer]

            ************* MSc THESIS AVAILABLE **************
            ***     DO NOT FORWARD TO ANY OTHER LISTS     ***
            *************************************************

The following thesis has been placed in the cs.huji.ac.il (132.65.16.10)
anonymous ftp. The file is igth.ps.gz (gzip compressed postscript) or
igth.ps.Z (compressed postscript). Ftp instructions follow the abstract.

         -----------------------------------------------------

      Unsupervised Learning of Cell Activities in the Associative
        Cortex of Behaving Monkeys, Using Hidden Markov Models

                                Itay Gat 
                    Institute of Computer Science and
                      Center for Neural Computation
                Hebrew University, Jerusalem 91904, Israel

                                ABSTRACT:

Hebb hypothesized in 1949 that the basic information processing unit in the
cortex is a cell-assembly which can act, briefly, as a closed system after
stimulation has ceased, and constitutes the simplest instance of a
representative process. This hypothesized cell-assembly may include thousands 
of cells in a highly interconnected network. The cell-assembly hypothesis 
shifts the focus from the single cell to the  complete network activity.
So far, there has been no general method for relating extracellular 
electrophysiological measured activity of neurons in the associative cortex to
the underlying network or the cell-assembly states. It is proposed here to 
model such data as a pair-wise correlated multivariate Poisson process. Based 
on these parameters, a Hidden Markov Model was computed. This modeling yielded
a temporal segmentation and labeling of the data into a sequence of states. 

The first hypothesis of this work was that a connection exists between the 
states of the model and the behavioral events of the animal. i.e. based on the
sequence of states of the model, the observed actions of the animal could be 
predicted. The second hypothesis was that a connection exists between the 
states defined above and the functional interaction between cells, i.e. the 
functional interaction between pairs of cells changes in different cognitive 
states.

The application of this approach was demonstrated for temporal segmentation of
the firing patterns, and for characterization of the cortical responses to 
external stimuli.This modeling was applied to 6 recording sessions of several 
single-unit recordings from behaving monkeys. At each session, 6-8 single-unit
spike trains were recorded simultaneously. Using the Hidden Markov Model, two
behavioral modes of the monkey  were significantly discriminated. The two 
behavioral modes  were  characterized by different firing patterns, as well as
by the level of coherency of their multi-unit firing activity. The result of 
the modeling showed a high degree of consistency, which implies that the model
succeeds in capturing a basic structure underlying the data. Significant 
changes were found in the temporal cross-correlation of the same pair of cells
in different states, indicating different functional connectivities of the 
small network being recorded. These changes suggest that the modeling captures
the activity of the network and that the states of the model can be related to
the cognitive states of the cortex.

         -----------------------------------------------------

                        FTP INSTRUCTIONS

unix> ftp cs.huji.ac.il (or 132.65.16.10)
Name: anonymous
Password: full_email_address
ftp> cd /pub/singer
ftp> binary
ftp> get igth.ps.Z
or
ftp> get igth.ps.gz
ftp> quit
unix> uncompress igth.ps.Z
or 
unix> gunzip igth.ps.gz
unix> lpr -P<printer-name> igth.ps


----------------------------------------------

Itay Gat
Institute of Computer Science
The Hebrew University, Givat-Ram
Jerusalem 91904, Israel
Email: itay at cs.huji.ac.il
Fax:   +972 2 58 5439