Artificial BioChemical Networks

摘要

Connectionist approaches to Artificial Intelligence are almost always based on Artificial Neural Networks. However, there is another route towards Parallel Distributed Processing, taking as its inspiration the intelligence displayed by single celled creatures called Protoctists (Protists). This is based on networks of interacting proteins. Such networks may be used in Pattern Recognition and Control tasks and are more flexible than most neuron models. In this paper they are demonstrated in Image Recognition applications and in Legged Robot control. They are trained using a Genetic Algorithm and Back Propagation.