Three Learning Paradigms

Machine learning is best understood through the lens of three core paradigms.

Supervised Learning

Trains on labelled examples. Evaluated with metrics like MAE, RMSE, accuracy, precision, recall, and F1.

  • Regression — Linear, Polynomial, Ridge, Lasso
  • Classification — Logistic Regression, Decision Trees, Random Forests, SVMs
  • Boosting — XGBoost, LightGBM, CatBoost

Unsupervised Learning

Finds hidden structure without labels.

  • Clustering — k-Means, DBSCAN, Hierarchical
  • Dimensionality Reduction — PCA, t-SNE, UMAP

Reinforcement Learning

Trains an agent via rewards and penalties. Key algorithms: Q-Learning, PPO, SAC.

"All models are wrong, but some are useful." — George Box

This guide maps the most important algorithms onto these paradigms with bias-variance trade-off notes and typical use-case pointers.