U of Toronto CRG-TR-90-3 announcement

[crg-tech-reports@cs.toronto.edu]

          DO NOT FORWARD TO OTHER NEWSGROUPS OR MAILING LISTS
          ***************************************************

The following technical report is now available.  If you'd like a copy please
send me your real mail address (omitting all other information from your
message).  Also, do not reply to the entire mailing list.

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

                    EXPERIMENTS ON DISCOVERING HIGH ORDER
                      FEATURES WITH MEAN FIELD MODULES

                  Conrad C. Galland & Geoffrey E. Hinton
       
                       Department of Computer Science
                           University of Toronto
                          Toronto, Canada M5S 1A4

                               CRG-TR-90-3


A new form of the deterministic Boltzmann machine (DBM) learning procedure is
presented which can efficiently train network modules to discriminate between
input vectors according to some criterion.  The new technique directly
utilizes the free energy of these "mean field modules" to represent the
probability that the criterion is met, the free energy being readily
manipulated by the learning procedure.  Although conventional deterministic
Boltzmann learning fails to extract the higher order feature of shift at a
network bottleneck, combining the new mean field modules with the mutual
information objective function rapidly produces modules that perfectly extract
this important higher order feature without direct external supervision.

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