@saachi27

Convexity, Logistic Loss, and the Overfitting Battle

May 12 · 6 min read · When moving from regression to classification, reusing the exact same optimization components without evaluating their underlying geometry can break your training pipeline. Today, I realized why the M

Logistic Regression and Decision Surfaces

Feb 3 · 5 min read · Up to this point, our entire pipeline has been focused on predicting continuous values—like housing prices—where the output can theoretically be any real number. But when the task shifts to classifica

Optimization Mechanics — Feature Scaling, Convergence Testing, and Polynomial Mapping

Jan 21 · 6 min read · When moving from univariate models to complex multivariate datasets, writing raw optimization loops is not enough. Without proper preprocessing and diagnosis, your model can suffer from exploding grad

Upgrading to Multiple Linear Regression

Jan 18 · 5 min read · Univariate linear regression is a great sandbox, but real-world datasets never rely on just a single feature. If you want to predict something complex—like house prices—you need to account for multipl

gradient_descent.py

Jan 15 · 5 min read · Deriving the cost function is only half the battle. To actually "learn," a model requires an optimization engine that shifts its parameters toward the lowest possible error without manual intervention