Notes on Mathematics & Data Science
Short pieces on the places where the two fields turn out to be the same conversation — where a modeling trick in machine learning is quietly doing linear algebra, or a piece of probability theory shows up wearing an optimization method's clothes. These aren't tutorials; I'm less interested in how to use a technique than in what's actually happening underneath it.
Hosted here: elapolak.github.io/ela-blog
Two Views on L2 Regularization
Ridge regression shows up everywhere in data science, usually justified with a shrug — "it just works." This piece traces why: the exact same penalty term turns out to be both a Bayesian MAP estimate under a Gaussian prior and a continuous, soft-shrinkage version of PCA, two completely unrelated mathematical stories arriving at the identical equation. Read the full piece →