Evolving Simple Symbolic Regression Models by Multi-objective Genetic Programming
M. Kommenda, G. K. Kronberger, M. Affenzeller, S. M. Winkler, B. Burlacu - Evolving Simple Symbolic Regression Models by Multi-objective Genetic Programming in Genetic Programming Theory and Practice XIII (Editors: Rick Riolo, William P. Worzel, Mark Kotanchek, Arthur Kordon) - Springer Gabler, 2016, pp. 1-19
In this chapter we examine how multi-objective genetic programming can be used to perform symbolic regression and compare its performance to single-objective genetic programming. Multi-objective optimization is implemented by using a slightly adapted version of NSGA-II, where the optimization objectives are the model’s prediction accuracy and its complexity. As the model complexity is explicitly
defined as an objective, the evolved symbolic regression models are simpler and more parsimonious when compared to models generated by a single-objective algorithm.
Furthermore, we define a new complexity measure that includes syntactical and semantic information about the model, while still being efficiently computed,
and demonstrate its performance on several benchmark problems. As a result of the multi-objective approach the appropriate model length and the functions included in the models are automatically determined without the necessity to specify them