Info-Evo: Using Information Geometry to Guide Evolutionary Program Learning
This work addresses the challenge of inefficient search in evolutionary algorithms for program synthesis, offering a domain-specific improvement.
The paper tackles the problem of guiding evolutionary program learning by introducing Info-Evo, a novel optimization strategy that uses natural gradient search with nonparametric Fisher information to direct evolutionary processes along 'shortest paths', resulting in improved efficiency in automated program learning.
A novel optimization strategy, Info-Evo, is described, in which natural gradient search using nonparametric Fisher information is used to provide ongoing guidance to an evolutionary learning algorithm, so that the evolutionary process preferentially moves in the directions identified as "shortest paths" according to the natural gradient. Some specifics regarding the application of this approach to automated program learning are reviewed, including a strategy for integrating Info-Evo into the MOSES program learning framework.