Path: utzoo!attcan!uunet!bu.edu!snorkelwacker!ira.uka.de!fauern!lan!kiss.informatik.tu-muenchen.de!li
From: li@kiss.informatik.tu-muenchen.de.informatik.tu-muenchen.dbp.de
Newsgroups: comp.ai.neural-nets
Subject: Learning parity function by backprob.
Message-ID: <4803@tuminfo1.lan.informatik.tu-muenchen.dbp.de>
Date: 9 Oct 90 17:17:33 GMT
Sender: news@lan.informatik.tu-muenchen.dbp.de
Reply-To: li@kiss.informatik.tu-muenchen.de.informatik.tu-muenchen.dbp.de ()
Organization: Inst. fuer Informatik, TU Muenchen, W. Germany
Lines: 11

Parity functions may be realized by NN with one hidden layer (a simple
solution was given in PDP-1). It is however a hard  problem to get such
solution by back-propagation algorithm. I was able to train a  NN with
backprog and some heuristcs to realize the P_4 (i.e. the parity
function of 4 bits vectors, P_2 is the XOR function). The P_5 seems, by
my experience, already to be too difficult to be learned by backprop,
no matter how many layers and neurons are used.  Does someone know
better results?

Thanks.
Xinzhi Li