作者: David Rosenbluth
DOI:
关键词:
摘要: The present invention is a node for network that combines Hopfield and Tank type neuron, having sigmoid transfer function, with nonmonotonic function such as parabolic to produce neural deterministic chaotic response suitable quickly globally solving optimizatioin problems avoiding local minima. can be included in completely connected single layer network. neuron operates continuously while the periodically prevent from getting stuck optimum solution. also area architecture where areas linked together hierarchy of neurons.