A constructive compound neural networks. II Application to artificial lifein a competitive environment

We have developed a new efficient neural network-based algorithm for Alife application in a competitive world whereby the effects of interactions amon g organisms are evaluated in a weak farm by exploiting the position of near est food elements into consideration but not the positions of the other com peting organisms. Two online learning algorithms, an instructive ASL (adapt ive supervised learning) and an evaluative feedback-oriented RL (reinforcem ent learning) algorithm developed have been tested in simulating Alife envi ronments with Various neural network algorithms. The constructive compound neural network algorithm FuzGa [15] guided by the ASL learning algorithm ha s proved to be most efficient among the methods experimented including the classical constructive cascaded CasCor algorithm of [18], [19] and the fixe d non-constructive fuzzy neural networks. Adopting an adaptively selected b est sequence of feedback action period Delta alpha: which we have found to be a decisive parameter in improving the network efficiency, the ASL-guided FuzGa had a performance of an averaged fitness value of 541.8 (standard de viation 48.8) as compared with 500(53.8) for ASL-guided CasCor and 489.2 (3 9.7) for RL-guided FuzGa. Our FuzGa algorithm has also outperformed the Cas Cor in time complexity by 31.1%. We have elucidated how the dimension less parameter food availability F-A representing the intensity of interactions among the organisms relates to a best sequence of the feedback action perio d Delta alpha and an optimal number of hidden neurons for the given configu ration of the networks. We confirm that the present solution successfully e valuates the effect of interactions at a larger F-A, reducing to an isolate d solution at a lower value of F-A. The simulation is carried out by thread functions of Java by ensuring the randomness of individual activities.