Lexicase Selection


Visualization of three lexicase selection events [2].

Lexicase selection is a recent parent selection technique originally developed by Lee Spector and Thomas Helmuth for solving program synthesis problems. It promotes solutions that perform well on uncommon sets of training cases. I developed epsilon-Lexicase selection, which extends this algorithm to continuous data [1]. We’ve done some work to understand lexicase and epsilon-lexicase selection using probabilistic and multi-objective analysis [2]. We’ve also extended it for many-objective optimization [3] and made a survival version for ensemble learning [4].

  1. La Cava, W., Spector, L., & Danai, K. (2016, July). Epsilon-lexicase selection for regression. In Proceedings of the 2016 on Genetic and Evolutionary Computation Conference (pp. 741-748). ACM. preprint

  2. La Cava, W., Helmuth, T., Spector, L., & Moore, J. H. (2018). A probabilistic and multi-objective analysis of lexicase selection and epsilon-lexicase selection. Evolutionary Computation, 1–28. link, arXiv

  3. La Cava, W., & Moore, J. H. (2018). An Analysis of epsilon-lexicase Selection for Large-scale Many-objective Optimization. GECCO 2018 Companion. link

  4. La Cava, W., & Moore, J. H. (2017). A general feature engineering wrapper for machine learning using ϵ-lexicase survival. European Conference on Genetic Programming, 80–95. https://doi.org/10.1007/978-3-319-55696-3_6. preprint