Linear Algebra · Linear algebra
SVD & PCA
Every matrix is rotate-stretch-rotate: A = UΣVᵀ, singular values, low-rank approximation (Eckart–Young), and principal component analysis for compressing and visualizing data, with image and finance examples.
Step 1 of 5 · Overview30 min read ·
Key idea
Prereqs: Matrix multiplication; eigenvalues; dot products. Next: Optimization for ML. Sources: Strang, *Introduction to Linear Algebra* ch. 7; Brunton & Kutz, *Data-Driven Science and Engineering* ch. 1; MIT OCW.Next: 1. The decomposition