<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<HTML><HEAD>
<META content="text/html; charset=koi8-r" http-equiv=Content-Type><BASE 
href="file://C:\Program Files\Common Files\Microsoft Shared\Stationery\">
<STYLE>BODY {
	BACKGROUND-POSITION: left top; BACKGROUND-REPEAT: no-repeat; COLOR: #000000; FONT-FAMILY: Arial; FONT-SIZE: 10pt
}
</STYLE>

<META content="MSHTML 5.00.2614.3500" name=GENERATOR></HEAD>
<BODY background=cid:002e01c2a873$7ed3a460$0100007f@podolsk.ru bgColor=#ffffff>
<DIV>\\<BR>Title: Parametrical Neural Network<BR>Authors: B.V.Kryzhanovsky, 
L.B.Litinskii, A.L.Mikaelyan (Institute of Optical <BR>Neural Technologies 
Russian Academy of Sciences)<BR>Comments: 16 pages, 2 Postscript figures, 
sumbitted to Neural Networks <BR>Report-no: <BR>Subj-class: Disordered Systems 
and Neural Networks<BR>\\<BR>The storage capacity of the Hopfield model is about 
15\% of the network size.<BR>It can be increased significantly in the 
Potts-glass model of the associative<BR>memory only. In this model neurons can 
be in more than two different states. We<BR>show that even greater storage 
capacity can be achieved in the parametrical<BR>neural network (PNN) that is 
based on the parametrical four-wave mixing process <BR>that is well-known in 
nonlinear optics. We<BR>present a uniform formalism allowing us to describe both 
PNN and the<BR>Potts-glass associative memory. To estimate the storage capacity 
we use the<BR>Chebyshev-Chernov statistical technique.&nbsp; 
<BR>\\<BR></DIV></BODY></HTML>
