Statistical Genetic Programming: The Role of Diversity

In the field of Genetic Programming (GP), there has been a growing interest in the effects of loss of genetic diversity, which causes the whole population prematurely converge to local optima. Improving diversity of the population is always an implicit goal of almost any basic genetic programming system. Most research in this area suggests a diversity measurement and controls this quantitative metric to maintain genetically diverse populations. This paper brief overviews of the measures used in Genetic Programming for diversity maintenance and promotion.

Journal of Intelligent and Robotic Systems, 2006

2003

Abstract This paper presents a study that evaluates the influence of the parallel genetic programming (GP) models in maintaining diversity in a population. The parallel models used are the cellular and the multipopulation one. Several measures of diversity are considered to gain a deeper understanding of the conditions under which the evolution of both models is successful. Three standard test problems are used to illustrate the different diversity measures and analyze their correlation with performance.

Lecture Notes in Computer Science, 2009

Population size is a critical parameter that affects the performance of an Evolutionary Computation model. A variable population size scheme is considered potentially beneficial to improve the quality of solutions and to accelerate fitness progression. In this contribution, we discuss the relationship between population size and the rate of evolution in Genetic Programming. We distinguish between the rate of fitness progression and the rate of genetic substitutions, which capture two different aspects of a GP evolutionary process. We suggest a new indicator for population size adjustment during an evolutionary process by measuring the rate of genetic substitutions. This provides a separate feedback channel for evolutionary process control, derived from concepts of population genetics. We observe that such a strategy can stabilize the rate of genetic substitutions and effectively accelerate fitness progression. A test with the Mackey-Glass time series prediction verifies our observations.

2009

• Papers developing techniques tested on small-scale problems include discussion of how to apply those techniques to real-world problems, while papers tackling real-world problems have employed techniques developed from theoretical work to gain insights.

2003

This paper presents an analysis of increased diversity in genetic programming. A selection strategy based on genetic lineages is used to increase genetic diversity. A genetic lineage is defined as the path from an individual to individuals which were created from its genetic material. The method is applied to three problem domains: Artificial Ant, Even-5-Parity and symbolic regression of the Binomial-3 function. We examine how increased diversity affects problems differently and draw conclusions about the types of diversity which are more important for each problem. Results indicate that diversity in the Ant problem helps to overcome deception, while elitism in combination with diversity is likely to benefit the Parity and regression problems.

1999

In this paper we investigate the influence of (a) the amount of variation generated in the genotype and (b) the depth of application of variation operators on the offspring fitness in genetic programming. Simulation results on three common test problems indicate that for certain features of the fitness distribution the location of the variation may play as important a role as the choice of the applied operators.

2002

Traditional GP randomly combines subtrees by applying crossover and mutation. There is a growing interest in methods that can control such recombination operations. In this study a new approach is presented for guiding the recombination process for GP. Our method is based on extracting the global information of the promising solutions that appear during the genetic search. The aim is to use this information to control the crossover operation afterwards.

Genetic and Evolutionary Computation, 2011

, except for brief excerpts in connection with reviews or scholarly analysis. Use in connection with any form of information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed is forbidden. The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject to proprietary rights.

Artificial Intelligence Review, 2002

Generalisation is one of the most important performance evaluation criteria for artificial learning systems. An increasing amount of research has recently concentrated on the robustness or generalisation ability of the programs evolved using Genetic Programming (GP). While some of these researchers report on the brittleness of the solutions evolved, some others propose methods of promoting robustness/generalisation. In this paper, a review of research on generalisation in GP and problems with brittleness of solutions produced by GP is presented. Also, a brief overview of several new methods promoting robustness/generalisation of the solutions produced by GP are presented.