Paper Announcement --- Learning in Boltzmann Trees

[Lawrence K. Saul]

FTP-host: archive.cis.ohio-state.edu
FTP-file: pub/neuroprose/saul.boltzmann.ps.Z

The file saul.boltzmann.ps.Z is now available for copying in the  
Neuroprose repository:

Learning in Boltzmann Trees (11 pages)
Lawrence Saul and Michael Jordan
Massachusetts Institute of Technology

ABSTRACT: We introduce a family of hierarchical Boltzmann machines  
that can be trained using standard gradient descent.  The networks  
can have one or more layers of hidden units, with tree-like  
connectivity.  We show how to implement the learning algorithm for  
these Boltzmann machines exactly, without resort to simulated or  
mean-field annealing. Stochastic averages are computed by the  
technique of decimation.  We present results on the problems of N-bit  
parity and the detection of hidden symmetries.


Lawrence Saul
lksaul at cmt6.mit.edu