Feed-forward network training using optimal input gains | IEEE Conference Publication | IEEE Xplore