Evolutionary Neurogenesis Research

Artificial neural networks (ANNs) are very useful modelling tools for a wide variety of problems. Unfortunately, the design of large-scale ANNs for complex problems can be an extremely difficult task. One promising approach to this scaling problem involves the growth of ANNs from a genotype program, such as set of rewrite rules. These genotypes provide a space-conserving recipe instead of a blueprint for the ANN, thus allowing the repetition of structures in the phenotype without a corresponding repetition in the genotype. Subjecting the phenotypes to an artificial selection process then paves the way for evolutionary algorithmic (EA) solutions to designing large ANNs.

Interesting Neurogenesis Papers

  1. Yao (1999) Overview of EAs+ANNs
  2. Large EA-ANN bibliography
  3. Boers & Sprinkhuizen-Kuyper(2001) The G2L System