[visarga]

Just 100 iterations, learning rate 0.03, activation tanh, regularization L2, rate 0.01. The network is 8,8,8 neurons per layer.