Genetic specification of recurrent neural networks: Initial thoughts

Computational Neuro-Genetic Modeling (CNGM) is discussed from the perspective of building Artificial Neural Network architectures starting with substantially pre-defined modules and processes (DNA-ANNs). This is equivalent to assuming that DNA code in a neuron can ultimately specify function, process and some level of data abstraction beyond the immediate role of genes to produce proteins or to regulate processes, and using that basis as a metaphor for DNA-ANNs. The potential advantages that might be derived from highly evolved, fine-grained hybrid genetic/connectionist systems, and some of the implementation challenges that they could present are discussed.