AIFAD

2004-04-18:  Minor speed improvement.

2004-03-24:  Fixed a bug.
             New way of computing entropy in gain ratio for subvariables
             under redundant structures: takes into account probability
             estimates from multinomial estimation, which may differ
             from a pure frequency based approach.
             Made Ristad's "Natural Law of Succession" the current default
             for multinomial estimation instead of the frequency based
             approach.  Experiments will still have to be performed to
             test its performance.

2004-03-22:  Intermediate release to fix a bug.  Changed computation
             of most frequent value again (dependent version).

• AIFAD improves decision tree learning by supporting significantly more complex kinds of data. This allows users to more conveniently describe the data they want to have learnt, which can improve accuracy and complexity of resulting models.
• AIFAD is free software published under the GPL.
• Learn more about this tool by reading its README as HTML or gzipped PostScript.
• Download the source distribution.
• Download the gzipped Linux-binary (statically linked).
• The theory behind this approach is explained in my submitted paper Using Algebraic Datatypes as Uniform Representation for Structured Data.
• For a full treatment of the theory behind you will have to wait for my upcoming PhD-thesis ;-)

This document was translated from L^AT_EX by H^EV^EA.