[comp.ai.neural-nets] Question on Backpercolation

mam@luise.informatik.uni-dortmund.de (Martin Mandischer) (05/13/91)

Hi, out there

A couple of weeks ago Mark Jurik announced his new algorithm called 
backpercolation. On small toy problems it seemed to be supirior to all the 
conventional training algorithms.
My questions now are: 

Are there experinces with this algorithm on lage scale problems ?
Has someone an C implementation of the algorithm which works for arbitary 
network architectures ?

Before we start implementing and playing around with it we like to have
some responses of the networld

+--------------------------------------------+
| Martin Mandischer                          |
| University of Dortmund                     | 
| Department of Computer Science VI          | 
| P.O. Box 500 500                           |
| W-4600 Dortmund                            |
| Tel.: 0231 - 7554737                       |
| e-mail: mam@yen.informatik.uni-dortmund.de |
+--------------------------------------------+