People have made great progress in the field of artificial neural networks. Many neural network models were proposed and studied mainly by computer simulations, but the number of models with exactly soluble dynamics i...People have made great progress in the field of artificial neural networks. Many neural network models were proposed and studied mainly by computer simulations, but the number of models with exactly soluble dynamics is up to now very limited. Explicit solutions for dynamics of the pseudoinverse neural network which is superior to the Hopfield model in both storage capacity and error-tolerance were presented by I. Kanter et al. with replica method. The layered pseudoinverse neural network model has also been solved on condition that the numbers of neurons and layers approach infirtity. However,展开更多
基金Project supported by the National Natural Science Foundation of China and the Chinese National Basic Research Project "Nonlinear Science".
文摘People have made great progress in the field of artificial neural networks. Many neural network models were proposed and studied mainly by computer simulations, but the number of models with exactly soluble dynamics is up to now very limited. Explicit solutions for dynamics of the pseudoinverse neural network which is superior to the Hopfield model in both storage capacity and error-tolerance were presented by I. Kanter et al. with replica method. The layered pseudoinverse neural network model has also been solved on condition that the numbers of neurons and layers approach infirtity. However,
