From: Convergence of batch gradient learning with smoothing regularization and adaptive momentum for neural networks
Input
Output
1
−1
0