A novel neural-network gradient optimization algorithm based on reinforcement learning | IEEE Conference Publication | IEEE Xplore