Forming dynamic, self-organizing neural networks by means of Assembler Encoding
Tomasz Praczyk · Bulletin of the Military University of Technology · 2011
assembler Encoding is a new neuro-evolutionary method. To date, it has been tested in such problems as: an optimization problem, a predator-prey problem, and in an inverted pendulum problem. In all the cases mentioned, assembler Encoding was used to create neural networks with constant, invariable architecture. To test whether assembler Encoding is able to form other types of neural networks, next experiments were carried out. In the experiments, the task of assembler Encoding was to form self-organizing, dynamic neural networks. The networks were tested in the predator-prey problem. To compare assembler Encoding with other method, in the experiments, a modified version of standard neuro-evolution was also applied. The results of the experiments are presented at the end of the paper.