Abstract

Neural Networks are a nontraditional method for performing computations. Neural Network architecture is a fine grain, parallel architecture with simple processors coupled at long range to the inputs. Figure 1 shows a simple picture of a layered neural network. Computation is performed by multiplying each input to every processor (node) by a weighting factor and then summing the products. The sum is then thresholded and an output is obtained from each node. The nonlinear thresholding operation provides the decision making mechanism.

© 1988 Optical Society of America