MOSES Meta-Optimizing Semantic Evolutionary Search

MOSES is a new approach to program evolution, based on representation-building and probabilistic modeling. Output (e.g. using datasets from computational biology, sentiment evaluation, agent control, etc.) is a compact and comprehensible program written in a simple Lisp-like mini-language.

English (United States)