rushter/MLAlgorithms

Covers supervised learning (linear/logistic regression, SVM, Random Forests, GBDT), unsupervised methods (K-Means, GMM, PCA, t-SNE), and neural architectures (MLP, CNN, RNN, LSTM) built exclusively with NumPy, SciPy, and Autograd—enabling algorithm inspection without framework abstractions. Designed for educational exploration with runnable examples and minimal dependencies, allowing developers to trace gradient computation and modify implementations directly for learning purposes.

10,960 stars. No commits in the last 6 months.