From mboxrd@z Thu Jan  1 00:00:00 1970
Received: (from weis@localhost) by pauillac.inria.fr (8.7.6/8.7.3) id NAA29095 for caml-redistribution; Tue, 7 Oct 1997 13:17:05 +0200 (MET DST)
Received: from nez-perce.inria.fr (nez-perce.inria.fr [192.93.2.78]) by pauillac.inria.fr (8.7.6/8.7.3) with ESMTP id UAA14070 for <caml-list@pauillac.inria.fr>; Mon, 6 Oct 1997 20:26:23 +0200 (MET DST)
Received: from cycle.Stanford.EDU (cycle.Stanford.EDU [171.65.76.133])
	by nez-perce.inria.fr (8.8.7/8.8.5) with ESMTP id UAA07132
	for <caml-list@inria.fr>; Mon, 6 Oct 1997 20:26:21 +0200 (MET DST)
Received: (from arc@localhost) by cycle.Stanford.EDU (8.8.7/8.6.6) id LAA21557; Mon, 6 Oct 1997 11:26:14 -0700 (PDT)
Date: Mon, 6 Oct 1997 11:26:14 -0700 (PDT)
From: Andrew Conway <arc@sequence.Stanford.EDU>
Message-Id: <199710061826.LAA21557@cycle.Stanford.EDU>
To: caml-list@inria.fr, vincent@cyberman.parc.anglia.ac.uk
Subject: Re:  Neural nets in Caml
Sender: weis


In response to the request for neural network software in caml/ocaml:

I have some brief neural network routines in ocaml in my general
library which is available from:

   http://sequence-www.stanford.edu/~arc/publib.html

A trivial test function to show backpropagation on a neural network
with 2 input nodes, 2 nodes in a hidden layer, and one output node
(hence [|2;2;1|] argument to create) is:

let selftest () =
    let net = create [|2;2;1|] in
    let teachdataset = [|
        ([|1.0;1.0|],[|0.0|]);
        ([|1.0;0.0|],[|1.0|]);
        ([|0.0;1.0|],[|1.0|]);
        ([|0.0;0.0|],[|0.0|])
    |] in
    let testdataset = teachdataset in
    polyteachtest 5000 net teachdataset testdataset;
    net
;;

Warning: I don't actually use neural networks, so I haven't tested it all
that much. I had an application that I thought could use them, wrote
the library, and then abandoned it.

Warning: I think that generally neural networks are the wrong way to solve
a problem.

There are lots of other things in the library, most with a one line
comment :-(

Regards,

Andrew.

-----------------

J'ai un bibliotech "neural network" (et autres choses) en ocaml dans 
	
   http://sequence-www.stanford.edu/~arc/publib.html

Malheureusement, il n'y a pas beaucoup de documentation. Et presque
tout est en Anglais. 

Amities,

Andrew.