CIPER – Constrained Inductive Polynomial Equation for Regression

Research areas: Machine Learning

__about_ciper
ciper was developed from a similar equation discovery tool named lagrange. lagrange was developed as a b.cs. thesis of a (not so senior 🙂 colleague ljupco todorovski. basic difference between ciper and lagrange is ciper is used for regression – a prediction of a predefined (continuous) variable with an polynomial equation, while lagrange finds all equations that are valid for a given dataset.
an important improvement of ciper over lagrange is in reducing search space of polynomial equations. it searches the space using beam search. the other space reducing method is in using sub- and super-polynomial constraints. for more information, check publications, please.
__publications
a paper from ecml 2004 conference, and a simple ciper manual.
__projects
ciper was developed in years 2003 and 2004 under project named cinq.
__download
you can download a ms windows 2000 version of ciper here. i hope it works with 98, NT, and XP as well. if not, please send mail. i am not really sure about a linux version, try running ciper using wine.
__comparable_methods
linear regression, model trees, regression trees, mars, other regression tools.
__check_also
minimal description length, least squares, beam search.

3. jun 2004