Meta-optimizing semantic evolutionary search

I present MOSES (meta-optimizing semantic evolutionary search), a new probabilistic modeling (estimation of distribution) approach to program evolution. Distributions are not estimated over the entire space of programs. Rather, a novel representation-building procedure that exploits domain knowledge is used to dynamically select program subspaces for estimation over. This leads to a system of demes consisting of alternative representations (i.e. program subspaces) that are maintained simultaneously and managed by the overall system. Application of MOSES to solve the artificial ant and hierarchically composed parity-multiplexer problems is described, with results showing superior performance. An analysis of the probabilistic models constructed shows that representation-building allows MOSES to exploit linkages in solving these problems. Categories and Subject Descriptors I.2.2 [Artificial Intelligence]: Automatic Programming – Program synthesis; I.2.8 [Artificial Intelligence]: Pr...

• Type: Conference
• Year: 2007
• Where: GECCO
• Authors: Moshe Looks
• Keywords: GECCO 2007, MOSES, Program Subspaces, Semantic Evolutionary Search