The Complete SPRLIB & ANNLIB

conj_pr

- perform the Polak-Ribiere variant of the conjugate gradient descent algorithm (LOCAL)

SYNOPSIS

int conj_pr (p, n, iter, verbose)

ARGUMENTS

double p[] Vector of all the trainable weights present in the network.
int n Total number of trainable weights.
long iter Total number of iterations.
long verbose If TRUE, this function will print the MSE during the iterations.

RETURNS

TRUE if an error was detected, FALSE if no error was detected; the adapted weights in p.

FUNCTION

Given a starting point p[1..n], Polak-Ribiere minimization is performed on a the mean squared error, using its gradient as calculated by back propagation. See also the explanation in learn_cgdes_ffnet.

NOTE

This function is a modified version of the frprmn() function found in NRinC.

SEE ALSO

learn_cgdes_ffnet, conj_fr

This document was generated using api2html on Thu Mar 5 09:00:00 MET DST 1998