Tetsuya MAESHIRO, Ken-nosuke WADA
Evolutionary System for the Computer Screening of the Coding Regions of Human Genome
Abstract:We present here an evolutionary system for the discrimination of the splicing sites of the human genome, whose population consists of neural networks. Most important of the parameters characterizing the networks are the learning algorithm, the duration of learning period, the number of hidden units, and the inheritance type. Three different simulations were run, and we discuss the results. Although several other simulations are necessary to be analyzed, we show that in general, even with the environment change, there is the predominance of short learning period, simple learning algorithm and Lamarckian type.