Machine Learning

Classical machine learning and real-world tabular applications with in-browser sklearn labs on ExamMaster.

Undergraduate depth: concept notes and, where they help, in-browser labs. Reading is free.

Lessons

Credit Risk ApplicationCredit default prediction is a classical tabular classification job — frame the costs, inspect the sample, then fit a sklearn baseline.
Customer Churn ApplicationChurn prediction is a classical tabular classification job where the business metric (catching leavers) beats raw accuracy.
Rent Pricing ApplicationRent estimation is a classical regression job on structured housing features — predict a number, explain the error in rupees.
Review Sentiment ApplicationTF-IDF plus a linear classifier is classical text ML — useful on short reviews before any neural language model.
BoostingBoosting stacks trees in sequence so each new model chases residual errors — powerful on tabular data, and easy to overfit if you never early-stop.
Confusion MatrixName the four cells of a binary classifier — TP, FP, TN, FN — then compute and heat-map them on an imbalanced hold-out in the browser.
Decision TreesGrow an axis-aligned decision tree with Gini/entropy impurity, then fit a shallow tree and paint its regions in the browser.
Evaluation MetricsAccuracy, precision/recall and the train–test gap are how you judge classical predictors — the metric has to match the decision cost.
k-means ClusteringGroup similar rows without labels: k-means assign-and-move, when the geometry fails, and a browser lab that fits and scatters clusters.
L1, L2 and the C DialPick the regularisation strength — lambda, alpha, or sklearn's flipped C — by scoring a few values on a held-out validation pile. L2 tames weights; L1 can delete.
Learning SetupLearning types, problem framing, and feature/label layout are the contract before any classical model — hold-out splits land in the next lesson.
Linear RegressionPredict a continuous target as a weighted sum of features — the first classical model, with an in-browser OLS lab.
Linearity AssumptionsLogistic regression as the classification baseline after a continuous line — assumptions, the sigmoid map, and a tiny sklearn lab. Regularisation now lives in its own topic.
Logistic RegressionTurn a linear score into a class probability with a sigmoid — the classification baseline after you have fitted a line for continuous targets.
PCAPrincipal components compress and plot high-dimensional tables by ranking orthogonal directions of variance — not labels.
Precision and RecallPrecision and recall measure two different error costs — false alarms versus missed cases — and almost always trade off.
Random ForestsBagging and random forests average many trees on bootstrap samples and random feature subsets — stabler than one deep tree, at the cost of a single readable path.
Recommender SystemsPredict what a person will like from a user–item table — content-based versus collaborative filtering, neighbourhood similarity, a popularity baseline, and cold start — not one…
RegularisationWhy huge weights chase noise, and how an L2 (ridge) penalty charges weight size so a coincidence no longer wins the training-loss contest.
ROC and AUCROC plots true-positive rate against false-positive rate across thresholds; AUC summarises the ranking quality of a scorer.
Train / Test SplitHold out train, validation, and test folds so fit, tune, and final score never share rows — then cut them with a seeded sklearn split in the browser.
Training LossA loss is one number for how wrong a model is on the rows it studied — from a single miss to the training-loss score the fit minimises.

Practise Machine Learning

Reading every chapter below is free and needs no account. Practice, mocks and progress live in the app.

Continue with Google — freeNo card, no trial. Works offline once installed.