Nitric oxide (NO) is a newly discovered neuronal messenger which transmits information in brain by way of diffusion. This phenomenon suggests a non-localized form of learning in computational neural network models. Based on a new dynamical description of single neuron learning, the authors demonstrate that NO diffusion can speed up the learning as well as reduce noise when a neuron is storing a pattern. Based on this idea they present the theory and application of a competitive learning algorithm that simulates pattern identification and classification in neural networks.
展开▼
