ml_basics

For fun, educational purposes and self-torture, I'm going to implement simple ML stuff in C. First is obviously MNIST. Let's see if I will continue afterwards. An idea would also be to switch to C++ add cuda support.

Roadmap:

• Mnist dataloader - check
• start implementation in Python first (only numpy)
• simple MLP nn architecture:
  • neurons, activations
  • forward pass
  • backward pass
  • training loop
  • monitoring / plotting the loss
  • evaluation
• Add Conv layers
• Run on Arduino ?