Session:

Evolutionary Computation and Neural Networks (ECNN)
Monday March 01, 2004, 17.15 – 17.35

Session Chair:

Paulo Cortez, Miguel Rocha

Paper Title:

Evolving Modular Neural Networks to Solve Challenging Control Problems

Author(s):

S. Doncieux, University Paris 6, France
J.A. Meyer, University Paris 6, France

Abstract:

This article describes ModNet, a framework devoted to the evolution of modular neural controllers that affords possibilities of bootstrapping the search for efficient solutions to challenging problems. Initial knowledge may be provided either as modules assigned to specific computations, or as an overall connectivity pattern describing how modules could be connected to each other or to the controller's inputs and outputs. These possibilities are used to automatically design neural networks that control respectively two complex dynamic systems: a cartpole and a lenticular blimp.