ann1_doc.txt

统计模式识别算法包

ANN1_DOC.TXT-- XTAL Version 5 Regression Package

                Comments regarding the ANN source code 


     The source code used for ANN1 regression is derived from the
software supplied with the book "Practical Neural Network Recipes in
C++" by Timothy Masters and published 1993 by Academic Press.  The
original source was written to run on an IBM PC but was modified so
that it could be compiled to run on a Sun Sparcstation.  It will run
the same as described in the book with two exceptions: (1) The escape
key will not interrupt training as in the IBM version.  (2) The new
version is case sensitive with respect to filenames since Unix
filenames are case sensitive but DOS is not.

     Beware when using this module standalone , since although it
behaves as described in the book it does not operate as it is used
with XTAL.  In particular, XTAL rescales the response variable so that
it falls in the range .1 to .9 as required by the original software.
The book is therefore absolutely necessary as a reference if using
ANN1 separate from the XTAL package.

     This code was compiled with the "g++" compiler from the Free
Software Foundation.  It does not compile correctly with the Sun
compiler available on the Sun Sparcstations here at the U of M.