GenNets: genetically programmed neural nets-using the genetic algorithm to train neural nets whose inputs and/or outputs vary in time

The author shows that the generic algorithm (GA) can be applied successfully to training nonconvergent networks, and presents some examples of their extraordinary behavioral versatility. He first gives a brief summary of the GA and the genetic programming of neural networks. He shows how GP techniques were used to evolve GenNets with specified operating conditions, and demonstrates some of the extraordinary capacities of time-dependent GenNets. He also makes a plea to the neural network research community to 'shift its sights upwards' by devoting more effort to thinking about 'dynamic' neural networks in general, and the theory of GenNet dynamics and 'evolvability' in particular. >