Training neural network with damped oscillation and maximized gradient function | IEEE Conference Publication | IEEE Xplore