This software package provides a library of evolutionary algorithms targetted at computational modelling. The library makes it easy to evolve sets of good cognitive models from one or more theories against one or more experimental constraints, and then explore their parameter values.
The library provides evolutionary algorithms, and supports either single constraint, multiple constraints combined into one, or multiple constraints with non-dominated sorting. These algorithms fall into six categories, depending on whether you take models from a single or multiple theories.
An example of using the library to evolve models for the 5-4 experiment uses four different theories (discrimination network, connectionist network, context and prototype) and sixty constraints, and is illustrated in a graphical demonstration program.
Download latest version.
Screenshot from included demo:
Publication relating to this code:
-
P.C.R. Lane and F. Gobet, 'Evolving non-dominated parameter sets for computational models from multiple experiments', Journal of Artificial General Intelligence, 4:1-30, 2013. https://doi.org/10.2478/jagi-2013-0001 (open access)