Most of the libraries are not longer maintained because libraries like Sklearn already provides efficacious APIs for most machine learning methods. Still it is a good thing to implement algorithms from scratch to get a better understanding of them.
- A library of various of network that implemented with Theano from scratch. It can provide some most basic details that higher-level frameworks like Keras cannot.
- I tried to implement several linear algorithms in Theano so that it can be accelerated on GPU.
- A variety of machine learning models and some interesting algorithms from Computational Intelligence.