|Copyright 1996-2001 by Donald R. Tveter, commercial use is prohibited. Short quotations are permitted if proper attribution is given. This material CAN be posted elsewhere on the net if the posted files are not altered in any way but please let me know where it is posted. The main location is: http://dontveter.com/bpr/bpr.html|
The idea of extracting rules from trained networks is very popular among the believers in classical symbol processing methods where the doctrine is that people use rules not networks (and not cases). Extracting rules also gives you some ability to understand the "reasoning" the network uses to produce its answers. Besides these reasons for trying to extract rules there is work that shows that the rules you get can give better results than the entire network. Extracting rules amounts to a really extreme pruning of the network.
There is an online paper by Shavlik on extracting rules from networks.