Open visual.html to observe the training process of this simple MLP network, and how it makes the samples linearly separable through nonlinear transformations.
Details:
1. Backward based on numpy;
1. Dimensions of hidden layers are all 2, for visualization.
requirements: numpy plotly