Lab sessions

The 13 lab sessions, what each one covers, the function each asks you to write, and the topic to study first.

Each lab has a demonstration part and a function for you to write. The topic linked below contains a program you can run, and in most cases a working version of the function. Read it, understand it, then write and test your own on the full class dataset, using load_data(n=None).

Session	Lab	Function to write	Study first
1	Environment setup and dataset exploration	`summarise_dataset(df)`	Topic 4
2	Exploratory data analysis report	A 5-page PDF, using `PdfPages`	Topic 4
3	Gradient descent from scratch: CGPA regression	Mini-batch GD, within 0.3 RMSE of scikit-learn	Topic 5
4	Logistic regression: binary and multinomial	`evaluate_classifier(model, X_train, y_train, X_val, y_val)`	Topic 6
5	Ridge, lasso and elastic net	`compare_regularisation(X_train, y_train, X_val, y_val, C_values)`	Topic 7
6	Decision trees: criteria, CCP pruning and depth sweep	`dt_report(X_train, y_train, X_val, y_val)`	Topic 9
7	Random forests: OOB, bagging, feature subsampling	`rf_vs_single_tree(X_train, y_train, X_val, y_val, n_estimators_list)`	Topic 10
8	XGBoost, LightGBM and SHAP	`boosting_benchmark(X_train, y_train, X_val, y_val)`	Topics 11 and 12
9	K-means, hierarchical and DBSCAN	`cluster_pipeline(X_scaled, k_range)`	Topics 13 and 14
10	Dimensionality reduction and anomaly detection	`dim_reduction_dashboard(X_scaled, y_labels, sample_n=1000)`	Topics 15 and 16
11	Data leakage, cross-validation and nested CV	`full_cv_report(estimator, X, y, param_grid, outer_k=5, inner_k=3)`	Topic 17
12	Full metrics suite: ROC, PR, calibration, McNemar	`model_evaluation_report()`	Topics 18 and 20
13	Modular pipeline, FastAPI, ONNX and MLflow	`end_to_end_check(model_path, data_path, n_batch=20)`	Topics 21, 23, 24 and 25

Labs 2 and 13, and the UMAP part of Lab 10, belong on your own computer. See How to study.