Differential evolution of constants in genetic programming improves efficacy and bloat

Authors:
Shreya Mukherjee;Margaret J. Eppstein
Affiliations:
University of Vermont, Burlington, VT, USA;University of Vermont, Burlington, VT, USA
Venue:
Proceedings of the 14th annual conference companion on Genetic and evolutionary computation
Year:
2012

Citing 3
Cited 1

Differential Evolution – A Simple and Efficient Heuristic for Global Optimization over Continuous Spaces

Journal of Global Optimization
Improving Gene Expression Programming Performance by Using Differential Evolution

ICMLA '07 Proceedings of the Sixth International Conference on Machine Learning and Applications
Using differential evolution for symbolic regression and numerical constant creation

Proceedings of the 10th annual conference on Genetic and evolutionary computation

Effects of constant optimization by nonlinear least squares minimization in symbolic regression

Proceedings of the 15th annual conference companion on Genetic and evolutionary computation

Quantified Score

Hi-index	0.00

Visualization

Abstract

We employ a variant of Differential Evolution (DE) for co-evolution of real coefficients in Genetic Programming (GP). This GP+DE method is applied to 30 randomly generated symbolic regression problems of varying difficulty. Expressions were evolved on sparsely sampled points, but were evaluated for accuracy using densely sampled points over much wider ranges of inputs. The GP+DE had successful runs on 25 of 30 problems, whereas GP using Ephemeral Random Constants succeeded on only 6 and the multi-objective GP Eureqa on only 18. Although nesting DE slows down each GP generation significantly, successful GP+DE runs required many fewer GP generations than the other methods and, in nearly all cases, the number of nodes in the best evolved trees were smaller in GP+DE than with the other GP methods.