The Role of Nearly Neutral Mutations in the Evolution of Dynamical Neural Networks

The evolution of continuous time recurrent neural networks is increasingly being employed to evolve nervous systems for autonomous agents. Nonetheless, the picture of populations engaged in hill-climbing rugged fitness landscapes poses a problem of becoming trapped on a local hilltop. Developments in evolutionary theory and molecular biology have pointed to the importance of selective neutrality. The neutral theory claims that the great majority of evolutionary changes are caused not by Darwinian selection but by random drift of selectively neutral or nearly neutral mutants. However, with a few exceptions neutrality has generally been ignored in artificial evolution. This paper addresses the distribution of fitness effects of new mutations when evolving dynamical systems and provides evidence of an improved evolutionary search process when incorporating nearly-neutral drift. This is one of the most fundamental problems in artificial evolution, because it lies at the heart of maintaining a constant-innovative property.

[1]  Inman Harvey,et al.  Evolutionary robotics: the Sussex approach , 1997, Robotics Auton. Syst..

[2]  Inman Harvey,et al.  Through the Labyrinth Evolution Finds a Way: A Silicon Ridge , 1996, ICES.

[3]  Explorations in Evolutionary Visualisation , 2002 .

[4]  Inman Harvey,et al.  Evolving visually guided robots , 1993 .

[5]  M. Shackleton,et al.  An investigation of redundant genotype-phenotype mappings and their role in evolutionary search , 2000, Proceedings of the 2000 Congress on Evolutionary Computation. CEC00 (Cat. No.00TH8512).

[6]  Lionel Barnett,et al.  Netcrawling-optimal evolutionary search with neutral networks , 2001, Proceedings of the 2001 Congress on Evolutionary Computation (IEEE Cat. No.01TH8546).

[7]  Randall D. Beer,et al.  A Dynamical Systems Perspective on Agent-Environment Interaction , 1995, Artif. Intell..

[8]  K. Holsinger.  The neutral theory of molecular evolution , 2004 .

[9]  Phil Husbands,et al.  Neutral Networks and Evolvability with Complex Genotype-Phenotype Mapping , 2001, ECAL.

[10]  A. Eyre-Walker,et al.  Estimating the distribution of fitness effects from DNA sequence data: Implications for the molecular clock , 2003, Proceedings of the National Academy of Sciences of the United States of America.

[11]  T. Smith,et al.  Neutrality and ruggedness in robot landscapes , 2002, Proceedings of the 2002 Congress on Evolutionary Computation. CEC'02 (Cat. No.02TH8600).

[12]  L. Barnett.  TANGLED WEBS Evolutionary Dynamics on Fitness Landscapes with Neutrality , 1997 .

[13]  P. Stadler,et al.  Neutral networks in protein space: a computational study based on knowledge-based potentials of mean force. , 1997, Folding & design.

[14]  P. Schuster,et al.  From sequences to shapes and back: a case study in RNA secondary structures , 1994, Proceedings of the Royal Society of London. Series B: Biological Sciences.

[15]  Randall D. Beer,et al.  Evolving Dynamical Neural Networks for Adaptive Behavior , 1992, Adapt. Behav..