Solving Uncompromising Problems With Lexicase Selection

Abstract

Abstract-We describe a broad class of problems, called "uncompromising problems," characterized by the requirement that solutions must perform optimally on each of many test cases. Many of the problems that have long motivated genetic programming research, including the automation of many traditional programming tasks, are uncompromising. We describe and analyze the recently proposed "lexicase" parent selection algorition and show that it can facilitate the solution of uncompromising problems by genetic programming. Unlike most traditional parent selection techniques, lexicase selection does not base selection on a fitness value that is aggregated over all test cases; rather, it considers test cases one at a time in random order. We present results comparing lexicase selection to more traditional parent selection methods, including standard tournament selection and implicit fitness sharing, on four uncompromising problems: finding terms in finite algebras, designing digital multipliers, counting words in files, and performing symbolic regression of the factorial function. We provide evidence that lexicase selection maintains higher levels of population diversity than other selection methods, which may partially explain its utility as a parent selection algorithm in the context of uncompromising problems.

Comparison of Semantic-aware Selection Methods in Genetic Programming

“…docfk employs f (k) to evaluate clustering result and choose the optimal k for the given data. Apart from these variants of doc, we run the conventional Koza-style GP (gp in the following), which employs tournament of size 7 in the selection phase, implicit fitness sharing (ifs, Section 3.1) also with tournament of size 7, lexicase selection (lex) [4] (Section 3.3) and doclex, the hybrid of doc and lex proposed in Section 3.4.…”

“…Recent experiments by Helmuth et al [4] demonstrate LEX's strengths in problem solving and diversity maintenance.…”

“…In this study, we experimentally compare the presumably oldest method of this type, Implicit Fitness Sharing (ifs, [19]) with two approaches proposed recently: discovery of objectives by clustering (doc, [8,12]) and lexicase selection (lex, [4]). Additionally, compared to [8], we consider a new variant of doc that employs a different clustering algorithm.…”

Adaptive Test Selection for Factorization-based Surrogate Fitness in Genetic Programming

“…The difficulties of tests are calculated from the current population and 346 K. Krawiec, P. Liskowski thus change with evolution, which can help escaping local minima and diversifies the population. Diversification maintenance was also the main motivation for the recent Lexicase selection algorithm [11], that avoids aggregating interaction outcomes altogether: in each selection act, a random permutation of tests is generated, and the program from the current population which passes the longest uninterrupted sequence of tests is selected. Lexicase proved very effective in a range of contexts and applications.…”

Lexicase selection outperforms previous strategies for incremental evolution of virtual creature controllers

“…Further details of the Lexicase algorithm are presented in the next section. Helmuth et al (2014) have shown that Lexicase is effective at solving challenging problems in GP. Although it was initially introduced for GP, Lexicase has the potential to be a very effective algorithm in evolutionary robotics where objectives relating to performance (e.g.…”