역전파 학습시 초기 가중치가 학습의 조기포기화에 미치는 영향 ( The Effect of Initial Weights on Premature Saturation in Back-Propagation Learning )

In this paper an approach for selection of convergence coefficient in a backpropagation learning rule is presented. This approach uses a stochastic automata learning rule for selection of the best coefficient in each step of the learning phase. This approach is applied to a nonlinear function approximation problem. Simulation results show that it gives faster convergence than the conventional and adaptive learning rate backpropagation rules.

N.O. Ezzati

Papers

Selection of convergence coefficient with automata learning rule