From linear regression to XGBoost, SHAP, discriminant analysis, and causal inference. 24 lessons covering the algorithms, diagnostic skills (bias-variance, learning curves), feature engineering, time-series, recommender systems, dimensionality reduction, and the practical mastery (calibration, hyperparameter tuning, model interpretation) that turn a textbook reader into a working ML engineer.
Supervised, unsupervised, reinforcement — the three pillars of ML.
Confusion matrices, ROC curves, cross-validation — measure what matters.
The diagnostic mental model behind every 'more data vs more capacity vs more regularization' decision.
Fit lines and curves to data. Your first ML algorithm.
Prevent overfitting with Ridge, Lasso, and ElasticNet. Geometric intuition for sparsity.
Binary classification with the sigmoid function.
Generative classifiers in the Bayes family — Naive Bayes plus LDA and QDA.
Classify by asking your neighbors. Simple, powerful, visual.
*The Bayes-Optimal Connection*
Find the widest street between classes. The kernel trick explained.
A flowchart that learns — splits data by asking yes/no questions.
Why many weak learners beat one strong one. Bagging, random forests, OOB error, and stacking the way Kaggle Grandmasters do.
The dominant algorithm in tabular ML. Stagewise additive modeling, learning rate, early stopping, and choosing between XGBoost / LightGBM / CatBoost.
Reliability diagrams, Brier score, ECE, Platt scaling, and isotonic regression — make predict_proba mean what it says.
Find natural groups in data without labels. K-Means, DBSCAN, and hierarchical with linkage and dendrograms.
Soft clustering with Gaussian mixtures. The EM algorithm, BIC for choosing k, and density estimation.
Isolation Forest, One-Class SVM, GMM-based, LOF — anomaly detection as a modeling task (where outliers ARE the signal).
Grid, Random, Bayesian (Optuna), Hyperband — the practical skill that separates ML engineers from notebook tinkerers.
Open the black box. SHAP, LIME, partial dependence, ICE, permutation importance — the toolkit the EU AI Act now requires.
Spotify's recommender lives in a 200-dimensional taste space — but the playlist UI is a 2D grid.
*Related Lessons Across Tracks*
*Related Lessons Across Tracks*
Open Netflix and stare at the homepage. Open Spotify and look at Discover Weekly. Open YouTube and watch "Up next" auto-queue. Open Amazon and read "Customers who bought…
A pharma company runs a regression and finds patients who took drug X recovered 12 days sooner.
12 interactive labs — hands-on exercises for this track
Predict house prices based on square footage — drag points and watch the best-fit line adjust
You're blindfolded on a hilly field trying to find the lowest valley — feel the slope and take steps downhill
You're a bank deciding who gets a loan — draw the line that separates approved from denied
You're a marketing team grouping customers by shopping behavior — watch the algorithm find natural groups
Find the widest road you can build between two neighborhoods — that's what SVM does
Walk down a decision tree step by step — see how every question narrows down to the final answer
Squish high-dimensional data into 2D while keeping similar points close — explore how t-SNE reveals clusters
Draw your own data points and watch the model find the best decision boundary in real time
Drag a threshold across 30 scored predictions and watch the ROC curve paint itself. Switch to Precision-Recall for imbalanced data.
Slide the λ knob and watch feature coefficients shrink — smoothly under Ridge, to exact zero under Lasso, mixed under ElasticNet.
Training vs validation scores over sample size. Three scenarios (underfit/good/overfit) each leave a different fingerprint.
Same dataset, 4 panels: single tree vs bagging vs AdaBoost vs stacking. Animate boosting rounds upweighting misclassified points.
600 questions across 24 modules — check how well you understood this track.