EvoGraph: A package for evolving graphs and networks

Dear GAers:

I am glad to announce that a package for evolving generic graphs and
networks using genetic programming is available for download.
 http://www.egr.msu.edu/~hujianju/evograph/evograph.htm

EvoGraph is a package for graph synthesis by genetic programming. It is
developed based on strongly-typed lilgp, originally developed at GARAGe, MSU
and later patched by Sean Luke with the strongly-typed feature. Basically,
EvoGraph provides a set of GP functions and terminals along with a C++ graph
library for evolving arbitrary graphs. The same technique is widely used in
developmental genetic programming for evolving electric circuits, bond
graphs, neural networks.

A benchmark problem: the wireless access point configuration problem is
introduced in this package. This problem requires simultaneous search of
both topology and parameters, the same as in circuit synthesis. And there
could be a series of variations of this problem and the problem is also
scalable in terms of problem size.

This package is motivated by the fact that there are few progress of
introducing benchmark problems for evaluating genetic programming
techniques. Especially and surprisingly, there is no such a benchmark
problem for topology synthesis, in which the most significant achievements
of genetic programming is made. A paper for precisely describing this
problem is expected for the coming GECCO.

This package is still in its starting stage and there is no complete or very
detailed documents. But it is easy to use and understand after reading
related reference papers.




Jianjun Hu (George)
Genetic Algorithm Research & Application Group (GARAGe)
Department of Computer Science & Engineering Michigan State University
 [EMAIL PROTECTED]
Web: www.egr.msu.edu/~hujianju
Phone: 517-355-3796(o)