Foundational Mathematics & Backpropagation

deep-learning

Handwritten derivations of computational graphs, chain rule applications in neural networks, and Jacobian matrices.

These notes cover the first-principles mathematical foundation of neural networks:

  • Multivariate Chain Rule: Tracking gradients through computational nodes.
  • Matrix Calculus: Jacobians, Hessians, and dimension preservation during backprop.
  • Layer Derivations: Fully connected forward and backward pass equations.