An advanced evolution should not repeat its past errors

Caroline Ravisé (1, 2) , Michèle Sebag (1)

1 LMS - Laboratoire de mécanique des solides
2 LRI - Laboratoire de Recherche en Informatique

A safe control of genetic evolution consists in preventing past errors of evolution from be­ing repeated. This could be done by keeping track of the history of evolution, but main­taining and exploiting the complete history is intractable. This paper investigates the use of machine learning (ML), in order to extract manageable information from this history. More pre­cisely, induction from examples of past trials and errors provides rules discriminating er­rors from successful trials. Such rules allow to a priori estimate the desirability of future trials; this knowledge can support powerful control strategies. SeveraI strategies of ML-based control are ap­plied to a genetic algorithm, and tested on the RoyaI Road, a GA-deceptive, and a com­binatorial optimization problem. Comparing mutation control with crossover control yields unexpected results.

Domaines: Mécanique des solides [physics.class-ph]

Caroline Ravisé, Michèle Sebag. An advanced evolution should not repeat its past errors. 13th International Conference on Machine Learning (ICML96), 1996, Strasbourg, France. pp.400-408. ⟨hal-00116421⟩