JNET is my own ML library built from scratch.

Dependencies

1. Numpy
2. Typing

Features

1. Steepest gradient descent optimization
2. Extensive typing in Python
3. Tanh activation layer
4. Mean-squared error loss

TODO:

1. Cross Entropy or another Loss function
2. Implement my own Tensor class (currently using NP arrays as Tensors)
3. Adding support for ReLU, Sigmoidal or Leaky activation layers
4. Research other optimization strategies than SGD