Software for Bayesian learning available

[Radford Neal]

                          Announcing Software for

                    BAYESIAN LEARNING FOR NEURAL NETWORKS

                     Radford Neal, University of Toronto

 Software for Bayesian learning of models based on multilayer perceptron 
 networks, using Markov chain Monte Carlo methods, is now available by 
 ftp.  This software implements the methods described in my Ph.D. thesis,
 "Bayesian Learning for Neural Networks".  Use of the software is free 
 for research and educational purposes.

 The software supports models for regression and classification problems
 based on networks with any number of hidden layers, using a wide variety
 of prior distributions for network parameters and hyperparameters.  The
 advantages of Bayesian learning include the automatic determination of 
 "regularization" parameters, without the need for a validation set, 
 avoidance of overfitting when using large networks, and quantification of 
 the uncertainty in predictions.  The software implements the Automatic 
 Relevance Determination (ARD) approach to handling inputs that may turn 
 out to be irrelevant (developed with David MacKay).  For problems and 
 networks of moderate size (eg, 200 training cases, 10 inputs, 20 hidden 
 units), full training (to the point where one can be reasonably sure that 
 the correct Bayesian answer has been found) typically takes several hours 
 to a day on our SGI machine.  However, quite good results, competitive 
 with other methods, are often obtained after training for under an hour.  
 (Of course, your machine may not be as fast as ours!)

 The software is written in ANSI C, and has been tested on SGI and Sun
 machines.  Full source code is included.

 Both the software and my thesis can be obtained by anonymous ftp, or via
 the World Wide Web, starting at my home page.  It is essential for you 
 to have read the thesis before trying to use the software.

 The URL of my home page is http://www.cs.toronto.edu/~radford.  If for
 some reason this doesn't work, you can get to the same place using the
 URL ftp://ftp.cs.toronto.edu/pub/radford/www/homepage.html.  From the
 home page, you will be able to get both the thesis and the software.

 To get the thesis and the software by anonymous ftp, use the host name
 ftp.cs.toronto.edu, or one of the addresses 128.100.3.6 or 128.100.1.105.
 After logging in as "anonymous", with your e-mail address as the password,
 change to directory pub/radford, make sure you are in "binary" mode, and
 get the files thesis.ps.Z and bnn.tar.Z.  The file bnn.doc contains just
 the documentation for the software, but this is included in bnn.tar.Z,
 so you will need it only if you need to read how to unpack a tar archive,
 or don't want to transfer the whole thing.  The files ending in .Z should
 be uncompressed with the "uncompress" command.  The thesis may be printed 
 on a Postscript printer, or viewed with ghostview.

 If you have any problems obtaining the thesis or the software, please
 contact me at one of the addresses below.

 ---------------------------------------------------------------------------
 Radford M. Neal                                      radford at cs.toronto.edu
 Dept. of Statistics and Dept. of Computer Science  radford at stat.toronto.edu
 University of Toronto                    http://www.cs.toronto.edu/~radford
 ---------------------------------------------------------------------------