Title:  Second-Order Methods in Boltzmann Learning:  
        An Application to Speechreading

Authors: Morten With Pedersen 
         Section for Digital Signal Processing 
         Department of Mathematical Modelling 
         Technical University of Denmark B321  
         DK-2800 Lyngby, DENMARK 
         mwp@imm.dtu.dk

         David G. Stork 
         Machine Learning and Perception Group
         Ricoh Silicon Valley              
         2882 Sand Hill Road Suite 115 
         Menlo Park, CA 94025-7022 USA  
         stork@crc.ricoh.com

Abstract:
We introduce second-order methods for training and 
pruning of general Boltzmann networks trained with
cross-entropy error. In particular, we derive the 
second derivatives for the entropic cost function. 
We illustrate pruning on  Boltzmann zippers, applied 
to real-world data --- a speechreading (lipreading) 
problem.

Submitted for NIPS*97, Denver Colorado, December 1997.