This directory contains documents and programs by Radford M. Neal.
My e-mail address is radford@cs.utoronto.ca. You can also get to
this information via the World Wide Web, starting the URL
http://www.cs.utoronto.ca/~radford
In fact, that's the preferred way, since I may not keep this README
file up-to-date. AND IN FACT, I HAVEN'T KEPT IT UP-TO-DATE!
Files ending in .ps.Z are compressed PostScript - THESE MUST BE
TRANSFERRED IN BINARY FTP MODE. These files can be uncompressed
on Unix machines using the 'uncompress' command, then viewed with
'ghostview' or printed with 'lpr', or whatever. For the larger
documents, you may find that the -s option of lpr is needed.
Short files containing the abstract and the table of contents for
papers may also be available. Some documents have a ".ref" file that
gives a reference to published papers based on the document.
NOTE: The text of my thesis is no longer available on-line (you can
still get the abstract). A revised version of the thesis with some
new material has now been published by Springer-Verlag. You should
be able to order it at 1-800-Springer or orders@springer-ny.com. The
ISBN is 0-387-94724-8. The price is $32.95 (US).
The documents in this directory are as follows:
slice Technical Report No. 9722 on "Markov chain Monte Carlo methods
based on `slicing' the density function", 27 pages.
mc-gp Technical Report No. 9702 on "Monte Carlo implementation
of Gaussian process models for Bayesian regression and
classification", 24 pages.
ws-fa Technical Report No. 9607 on "Factor Analysis Using
Delta-Rule Wake-Sleep Learning", 23 pages.
over Technical Report No. 9508 on "Suppressing Random Walks in
Markov Chain Monte Carlo Using Ordered Overrelaxation", 22 pages.
ttrans Technical Report No. 9421 on "Sampling from multimodal
distributions using tempered transitions", 22 pages.
pin Technical report CRG-TR-94-1 on "Priors for infinite
networks", 22 pages.
review Technical report CRG-TR-93-1 on "Probabilistic inference
using Markov chain Monte Carlo methods", 144 pages.
emk Paper with Geoffrey Hinton on "A view of the EM algorithm that
justifies incremental, sparse, and other variants", 14 pages.
bbp Technical report CRG-TR-92-1 on "Bayesian training of
backpropagation networks by the hybrid Monte Carlo method",
21 pages.
bmm Technical report CRG-TR-91-2 on "Bayesian mixture modeling
by Monte Carlo simulation", 23 pages.
sff Technical report CRG-TR-90-7 on "Learning stochastic
feedforward networks", 34 pages.
taxc Unpublished manuscript on "The computational complexity of
taxonomic inference", December 1989, 18 pages.
The programs in this directory are as follows:
fbm.* These files contain versions of my software for
flexible Bayesian modeling, including neural network
and Gaussian process models. You're better off getting
them via my Web page.
bnn.* These files contain some old version of Bayesian neural
network software.
ws-fa.tar Programs implementing wake-sleep learning of factor
analysis models.
lowp_ac.shar Shell archive of software for fast arithmetic coding
using low-precision arithmetic, which can be done quickly
without multiplies and divides.
lowp_ac.doc Just the documenation for the above software (which is
included in lowp_ac.shar).
cacm_ac.shar Shell archive of adaptive arithmetic coding software
from Witten, Neal & Cleary, Communications of the ACM,
vol. 30, no. 6 (June 1987).