This project aim is getting better understanding of how NNs work inside. I plan to add here as many features as I can, and don't care too much about optimization.
- Add more optimizers(currently there is only hard-coded gradient descent)
- Add more layers(now only dense layer and activations)
- Add callbacks
- Make non-sequential architectures available?(I have no idea how to implement it, this will require writing some graph-related stuff, which I am not good at)