Lightweight and fast python implementation for training neural networks.
2-layer NN, 256 hidden units, mean square error, learning rate = 0.001, batch size = 64
2-layer NN, 512 hidden units, mean square error, learning rate = 0.001, batch size = 64
3-layer NN, 384 hidden units, mean square error, learning rate = 0.01, batch size = 64
4-layer NN, 512 hidden units, mean square error, learning rate = 0.01, batch size = 64