Neural systems: how the artificial version models itself after nature

The modeling of biological neurons is discussed, which leads to a system of nonlinear differential equations with several state variables for each neuron; this would be untenable in a computational application. A simple mathematical model is obtained retaining some essentials of real dynamic behavior. In order to make the network perform any task, it should be trained or it has to learn to solve the problem. The learning is inherent in biological systems. The learning procedures, supervised and unsupervised learning, are described. Current research on biological neural networks is discussed.<
>