Diploidy for evolving neural networks
Cara L. Reedy · Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2018
Genetic algorithms and artificial neural networks are two widely-used techniques that can be combined with each other to produce evolved neural networks. Some research has also looked at the use of diploidy in genetic algorithms for possible advantages over the haploid genetic representation usually used, most notably in the form of better adaptation to changing environments. This paper proposes a diploid representation for evolving neural networks, used in an agent-based simulation. Two versions of the diploid representation were tested with a haploid version in one static and two changing environments. All three genetic types performed differently in different environments.