Forming Neural Networks by Means of Assembler Encoding–Preliminary Report
Tomasz Praczyk · Intelligent Automation & Soft Computing · 2011
Abstract This paper presents anew neural network encoding method called Assembler Encoding. The method was tentatively tested in the predator-prey problem. In the experiments, neural networks created by means Assembler Encoding, were used to control predators whose task was to capture a fast moving prey behaving according to a simple strategy. To compare Assembler Encoding with another neural network encoding method, in the experiments, a co-evolutionary version of simple connectivity matrix was also applied. The results of the experiments are included in the final part of the paper.