Reconstructing signal from fiber-optic measuring system with non-linear perceptron

摘要

A computer model of the feed-forward neural network with the hidden layer isdeveloped to reconstruct physical field investigated by the fiber-opticmeasuring system. The Gaussian distributions of some physical quantity areselected as learning patterns. Neural network is learned by errorback-propagation using the conjugate gradient and coordinate descentminimization of deviation. Learned neural network reconstructs thetwo-dimensional scalar physical field with distribution having one or twoGaussian peaks.