Improving gradient-based learning algorithms for large scale feedforward networks | IEEE Conference Publication | IEEE Xplore