We are interested in learning programs for multiple related tasks given only a few training examples per task. Since the program for a single task is underdetermined by its data, we introduce a nonparametric hierarchical Bayesian prior over programs which shares statistical strength across multiple tasks. The key challenge is to parametrize this multi-task sharing. For this, we introduce a new representation of programs based on combinatory logic and provide an MCMC algorithm that can perform safe program transformations on this representation to reveal shared inter-program substructures. 1.

Abstract Welcome to genetic programming, where the forces of nature are used to automatically evolve computer programs. We give a flavour of where GP has been successfully applied (it is far too wide an area to cover everything) and interesting current and future research but start with a tutorial of how to get started and finish with common pitfalls to avoid. 1