LEARNABLE EVOLUTION MODEL: Evolutionary Processes Guided by Machine Learning

  • Authors:

  • Ryszard S. Michalski

  • Affiliations:

  • Machine Learning and Inference Laboratory, George Mason University, Fairfax, VA and Institute of Computer Science, Polish Academy of Sciences, Warsaw, Poland. michalski@gmu.edu

  • Venue:
  • Machine Learning - Special issue on multistrategy learning
  • Year:
  • 2000

Quantified Score

Hi-index 0.00

Visualization

Abstract

A new class of evolutionary computation processes ispresented, called Learnable Evolution Model or LEM. Incontrast to Darwinian-type evolution that relies on mutation,recombination, and selection operators, LEM employs machine learningto generate new populations. Specifically, in Machine Learningmode, a learning system seeks reasons why certain individuals in apopulation (or a collection of past populations) are superior toothers in performing a designated class of tasks. These reasons,expressed as inductive hypotheses, are used to generate newpopulations. A remarkable property of LEM is that it is capable ofquantum leaps (“insight jumps”) of the fitness function, unlikeDarwinian-type evolution that typically proceeds through numerousslight improvements. In our early experimental studies, LEMsignificantly outperformed evolutionary computation methods used inthe experiments, sometimes achieving speed-ups of two or more ordersof magnitude in terms of the number of evolutionary steps. LEM has apotential for a wide range of applications, in particular, in suchdomains as complex optimization or search problems, engineeringdesign, drug design, evolvable hardware, software engineering,economics, data mining, and automatic programming.