Implementations
A compact collection of fundamental learning systems, rebuilt from first principles. Each case study follows the mechanism, the implementation decisions, and the evidence used to verify it.
Algorithms Paradigms
Gaussian Mixture Models
Soft clustering via EM. Points can belong to multiple groups, which makes more sense than hard boundaries for a lot of real data.
2026
Gradient Boosting Machine
What gradient boosting actually does — one shallow tree at a time, each fitting the errors the previous trees left behind.
2026
Multi-Layer Perceptron
A basic neural network without any framework magic. Forward pass, backprop, weight updates — all written out in NumPy.
2026