Fast training analog approximator on the basis of Legendre polynomials

摘要: In a number of applications the approximation or interpolation certain weakly (when every subsequent term power series expansion is much less than previous one) nonlinear dependencies d(x), where x an arbitrary signal in time, demanded. The example problem cancellation distortion high precision analog engineering. such cases it seems to be reasonable use polynomial (interpolation) devices. this paper neural network based devices, performing operations interpolation, are described. schemes and working characteristics breadboard on radio components presented. Legendre polynomials were offered as basis functions for significant increasing speed approximator training. scheme synthesizer was also suggested.