Version française
Home     About     Download     Resources     Contact us    

This site is updated infrequently. For up-to-date information, please visit the new OCaml website at

Browse thread
[Caml-list] genetic programming library
[ Home ] [ Index: by date | by threads ]
[ Search: ]

[ Message by date: previous | next ] [ Message in thread: previous | next ] [ Thread: previous | next ]
Date: 2001-11-08 (15:13)
From: Markus Mottl <markus@o...>
Subject: Re: [Caml-list] genetic programming library
On Wed, 07 Nov 2001, wrote:
> Has anyone worked on a genetic programming library for OCaml? I am
> working on one and hate to duplicate work.

It's not quite genetic programming that I'm working on, but I have a
framework that allows you to plug in arbitrary heuristics for generating
syntactically correct programs, evaluating them and giving feedback to
the generation algorithm about the "quality" of the solution. Adding a
GP-like heuristics to the already existing program generation strategies
(complete enumeration of search space + random sampling) shouldn't be
too difficult.

The system implementation still lacks many convenience features (even
worse: documentation ;), but it might be a suitable testbed for what
you want to do. There is also a complete application that can be run on
C4.5 data (only for numeric regression, not for classification purposes
or on discrete data).

The implemented example "language" allows symbolic regression, though
at a somewhat more sophisticated level than you'll normally find in
literature: it uses efficient methods for estimating constants in
multivariate polynomials rather than just "guessing" them.

Right now I have to work on something else, which might at some later
(much later) point be combined with the previous system. If enough
people show serious interest in trying it, or better, contributing code,
I _might_ consider putting it online.

Markus Mottl

Markus Mottl                                   
Austrian Research Institute
for Artificial Intelligence        
Bug reports:  FAQ:
To unsubscribe, mail  Archives: