This repo contains machine learning algorithms coded from scratch. These implementations are not designed for high-performance, the focus lies on coding a generic implementation only with necessary code.
There are two blog posts about the code and the machine learning algorithms as whole. You can find the gradient boosting one here and the one about regression trees here.
This work was originally by Andre Bleier. I thought that it was a good idea but found that it had a bug that meant that the predictions were wrong. However, this was easily fixed. I hope to made some additions to the code in the future.