A Modified Compact Genetic Algorithm For The Intrinsic Evolution Of Continuous Time Recurrent Neural Networks

In the past, we have extrinsically evolved continuous time recurrent neural networks (CTRNNs) to control physical processes. Currently, we are seeking to create intrinsic CTRNN devices that combine a hardware genetic algorithm engine on the same chip with reconfigurable analog VLSI neurons. A necessary step in this process is to identify a genetic algorithm that is both amenable to hardware implementation and is sufficiently powerful to effectively search CTRNN spaces. In this paper, we will propose and test several variations of the compact genetic algorithm (CGA) for searching these spaces. We will then benchmark the best variant using the De Jong functions, outline a hardware implementation, and discuss future plans to develop an integrated evolvable hardware device controller.