Samarth Negi

Terence Parr and Jeremy Howard. All the matrix calculus behind neural network training, built up from scalar derivatives.

Patient, first-principles derivations of statistics and machine learning ideas. A model for how to write technical notes.