PENNSYLVANIA UNIV PHILADELPHIA
Pagination or Media Count:
A NEURAL NETWORK is a system of interconnected, excitable, neuronlike units. Units can be activated by inputs from other units in the network, by external inputs, or by intrinsic processes. Based on these inputs, each unit generates an output that is transmitted to all units to which it projects. The efficacy of connections in transmitting inputs and outputs can be modified, usually in an activity-dependent manner, and various rules have been proposed to govern these synaptic modifications. The most widely used rule the Hebb rule modifies connection efficacies according to the correlation of the excitation in the two connected units. Neural networks can act to generate, transform, or recognize patterns of activity, to store information, and to perform learning. Networks can be simulated on computers or directly implemented in hardware such as passive electrical circuits, VLSI chips, or optical components.