SGI MLC++ Utilities (NT Edition) README file.           November 1997
Machine Learning Library in C++.
http://www.sgi.com/tech/mlc

Keywords: machine learning, accuracy estimation, cross-validation,
          bootstrap, ID3, decision trees, decision graphs, naive-bayes,
          decision tables, majority, induction algorithms, classifiers,
          categorizers, general logic diagrams, instance-based algorithms,
          discretization, lazy learning, bagging, MineSet.

MLC++ Team: The MLC++ team is continuing to work on research
  and development of machine learning techniques for data mining.
  The team is now working at Silicon Graphics, as part of the
  MineSet project for data mining and visualization:
      http://mineset.sgi.com.
  Core team members include: Ronny Kohavi (manager)
                        Cliff Brunk
                        Alex Kozlov
                        Clay Kunz
                        Dan Sommerfield
                        Eric Eros (contractor).

Questions or help requests related to the utilities should be
    addressed to mlc@postofc.corp.sgi.com 

Please see the MLC++ home page first:
    http://www.sgi.com/tech/mlc

______________________________________________________________________
Quick starter:

The MLC++ utilities are accessible through our web page
   http://www.sgi.com/tech/mlc

The NT version of the MLC++ utilities are stored in a .zip archive,
and can be uncompressed by pkunzip as well as a number of other
common utilities.  Assuming pkunzip, follow this procedure to
unpack the archive:

1. Make a directory to hold the utilities.  We'll call this <mlc>

2. Place the archive file in <mlc>

3. Execute pkunzip <archive> within the <mlc> directory, or use
   the windows version of pkunzip to unpack the files.

The documentation is in utils.ps and it is currently OUT OF DATE.
Expect an update in the next month.

Environment variables:
The MLC++ utilities make heavy use of environment variables while
running.  Environment variables may be set from the NT command
prompt as follows:

  set <name>=<value>

You can check the value of a variable as follows:

  echo %<name>%

The following environment variables need to be set to get
MLC++ to run:
MLCDIR must be set to the directory where the utilities are installed.
MLCPATH to the directory where the databases are stored.  Multiple
   directories may be specified through this variable and separated
   by colons.

All directories specified in MLCDIR or MLCPATH must use FORWARD
(unix-style) slashes.  Drive letters should be specified as
follows:
     //c/
where c can be replaced by the drive letter of your choice.

Example:
   If you installed the utilities in C:\MLC, you would execute
   the following to set MLCDIR and MLCPATH:

   set MLCDIR=//c/MLC/db
   set MLCPATH=.://c/MLC/db

Note:
   This method for specifying path names is identical to that
   supported by the GNU-Win32 project
   (see http://www.cygnus.com for details), except that we
   do not support volume mounting.

File Formats:
   MLC++ can read ASCII files in both UNIX and DOS formats
   seamlessly.  The sample data files provided with the kit
   are actually in UNIX format.
   MLC++ will always produce DOS format text files when running
   under NT.

If you have not "registered" through the web page, please
do so at http://www.sgi.com/tech/mlc/mail.html

Databases in the MLC++ format, which is very similar to C4.5 format
  can be found in http://www.sgi.com/tech/mlc/db/
  Most datafiles are converted from the repository at UC Irvine.
______________________________________________________________________