Faster Convergence with Lexicase Selection in Tree-based Automated Machine Learning

TL;DR

This paper demonstrates that using lexicase selection in TPOT accelerates convergence in evolving machine learning pipelines compared to NSGA-II, with analysis of search space exploration.

Contribution

It provides the first comparative study of lexicase selection versus NSGA-II in automated machine learning pipeline evolution.

Findings

Lexicase selection leads to faster convergence in TPOT.

Lexicase explores the search space more efficiently.

Experimental results on multiple datasets support these conclusions.

Abstract

In many evolutionary computation systems, parent selection methods can affect, among other things, convergence to a solution. In this paper, we present a study comparing the role of two commonly used parent selection methods in evolving machine learning pipelines in an automated machine learning system called Tree-based Pipeline Optimization Tool (TPOT). Specifically, we demonstrate, using experiments on multiple datasets, that lexicase selection leads to significantly faster convergence as compared to NSGA-II in TPOT. We also compare the exploration of parts of the search space by these selection methods using a trie data structure that contains information about the pipelines explored in a particular run.