Towards the Automatic Programming of NEPs: A First Case Study
This work shows the first results of our platform for the automatic design of NEPs to solve specific tasks. The platform is based on a genetic programming algorithm that we have proposed earlier. It uses Christiansen grammars to exclude individuals with either syntactic or semantic mistakes. The fitness function required by the genetic engine, usually invokes a simulator of the model under consideration. In this work we use jNEP, a Java simulator for NEPs developed by our research group. We have chosen a non trivial problem borrowed from a NEP that applies context free rules for simulating pushdown automata: the rotation of the strings until finding the symbol to which the rule will be applied, which is one of the three steps this NEP takes. We have found some interesting solutions.
KeywordsContext Free Grammar Grammatical Evolution Terminal Symbol Trivial Problem Attribute Grammar
Unable to display preview. Download preview PDF.