Departmental Bulletin Paper ダイナミックバイナリーニューラルネットの学習と安定化

佐藤, 龍直

A dynamic binary neural network is a simple two-layer network with a delayed feedback and is able to generate various binary periodic orbits. The network is characterized by the signum activationfunction, ternary connection parameters, and integer threshold parameters. The ternary connection brings benefits to network hardware and to computation costs in numerical analysis.The dynamics is simplified into a digital return map on a set of lattice points. We investigate the relation between sparsity of network connection and stability of a target periodic orbit. In order to stabilize a desired binary periodic orbit, we introdece some methods algorithm uses Each individual is evaluated by some feature quantities that characterize the stability of the periodic orbit. Applying the algorithm to a class of periodic orbits that are applicable to control signals of switching power converters, the usefulness of sparsification in stabilization of desired periodicorbit is confirmed.Key Words : Dynamic binary neural networks, Stabilization, Feature quantities

Number of accesses :  

Other information