Domain 3: Classical Machine Learning
Supervised and unsupervised learning, boosting, clustering, t-SNE and UMAP, feature engineering, evaluation, imbalanced data and SHAP, with runnable code.
Unlock all 11 lessons with Unrolled Pro
Full production code, quantitative capacity math, failure recovery playbooks, and interview coaching.
Lessons
11 lessonsSupervised Learning — Learning From Labelled Examples
"Will this student quit?", "How long will this answer take to generate?", "Is this message spam?" — each is a question with known answers in historical data.
Three Classic Classifiers — k-Nearest Neighbours, Naive Bayes and Support Vector Machines
These three methods are built on three different ideas: similarity (kNN: you're like your neighbours), probability (Naive Bayes: which class most likely generated these features?
Boosting — From AdaBoost to Gradient Boosting, XGBoost, LightGBM, CatBoost and Stacking
Random forests average many independent deep trees to reduce variance. Boosting does the opposite: it builds small trees sequentially, each one correcting the errors of the ensemble so far, reducing bias.
Unsupervised Learning — Finding Structure Without Labels
Labels are expensive: someone has to read every student question to tag its topic.
Gaussian Mixtures and EM, t-SNE and UMAP — Soft Clusters and Honest Pictures of High-Dimensional Data
k-means gives hard assignments and assumes round clusters of similar size.
ML Problem Families — Choosing the Right Shape for the Problem
The hardest step in applied ML is often not the model — it's deciding what kind of problem you're solving. "Help students learn faster" could be classification (will they quit?), ranking (which lesson first?
Feature Engineering — Turning Raw Data Into Model Inputs
Real data arrives as logs: some fields missing, plan names as text, timestamps, IDs. A model only sees numbers.
ML Evaluation — Measuring Models Without Fooling Yourself
A metric is a promise about the future: "this model will behave like this on new students".
Imbalanced Data, Hyperparameter Search and SHAP — Making Tabular Models Work and Explaining Them
Rare positives (churn, fraud, failing students) break naive training and evaluation; hyperparameters decide whether a model is mediocre or excellent, and the search strategy decides how much compute that costs; and explanations decide whether people trust and correctly use a model.
Classical ML Capstone — Churn Prediction, Price Prediction, a Recommender, and the ML Interview
Knowing algorithms isn't the same as delivering a model that helps.
Bridge: When Hand-Made Features Stop Scaling
Atlas's classic models work on tidy tables — but they can't read a question's text, an image or a sequence without someone hand-crafting features.