We have updated GPLAB with the following features:
* Better graphics
* Added tracking of pareto front, which enhances
somewhat the current accuracy_complexity function
* Added possibility to switch off use of stop condition.
* Added possibility to calculate population diversity.
* Added possibility to do crossvalidation (separate
training and test set). This is enabled by providing
* Added method for drawing generated trees
* Possible to specify different genetic types. Currently
two genetic types are included: The normal GPLAB tree,
but also an type containing two or more trees. In the
latter case, when doing crossover between two
individuals, the first tree of individual A will be mixed
with the first tree of individual B, the second tree with
the second tree and so forth. We believe it is possible
to exploit this generalization to support arbitrary
* All demos should work (at least not crash) after these
changes. Comments should also reflect the actual
function of each method.
These features were implemented by SINTEF
(www.sintef.no). The work was sponsored by Hydro
(www.hydro.com). Please contact email@example.com for
questions and comments.