Abstract
It is shown how first-order polynomial correlation procedures can measure components of the regression function of a nonlinear network. The components can be determined independently if Hermite polynomials are used in the correlation procedure.

This publication has 0 references indexed in Scilit: