论文部分内容阅读
Echo state network(ESN)is a classical reservoir computing model,but the performances of ESN vary greatly among different realizations.This paper proposes a pre-adapting mechanism-intrinsic plasticity(IP)rule to improve the ESNs performance stability.The IP rule is an adjusting mechanism in biological neurons.It is considered the root to keep the biological neural network stable.The IP mechanism is imitated from a mathematical view.Also,the effects of the IP rules with and without restraints to the performance of ESNs are investigated through two experiments.The results show that the IP rules with restraints can improve the performance stability of ESN.Some further experimental studies reveal the cause why the IP rule with restraints improves the performance stability of ESN successfully.