Version v0.1.0

Here we go.

1. Usage of gonum for matrix calculations;
2. Fully-connected layer without bias;
3. ReLU layer;
4. Pooling layer. Only 'max' supported as pooling function currently;
5. Convolution layer without bias;
6. Gradient descent with inertia as backpropagation function for convolution and fully-connected layers;
7. Conv2D, Pool2D, ZeroPadding, Im2Col, Rotate, Reshape, ControursPadding, Flatten functions;
8. Training process;
9. Some simple examples for MLP, CNN;
10. Activation functions and its derivatives: tanh, sigmoid, atan, softplus, gaussian;
11. Graphivz for fully-connected layer;
12. Pretty prints on layers;
13. Import/Export JSON files.

Future and current works are here: https://github.com/LdDl/cnns/blob/master/ROADMAP.md