AE03 — Machine Learning Foundations

Build reliable baselines, select data representations, train models, and evaluate them correctly offline. Projects may use tabular data, text classification, or simple retrieval.

Tư vấn học phần này Lịch khai giảng

Prerequisites

AE02 competency and basic Python, NumPy, and pandas.

Outcomes

A reproducible offline pipeline, benchmark, error report, and model artifact. Learners can explain which data was used for fitting, tuning, and testing.

Audience

Software engineers moving into ML and analysts building predictive models.

Core capability

Build and evaluate ML models.

Syllabus

16 sessions in 8 weeks.

Week	First session	Second session
1	S1: Turn questions into supervised or unsupervised learning problems, then define targets and baselines	S2: Train, validation, and test sets, including time-based and group-based splits and data leakage
2	S3: Linear regression, fitting, residuals, and prediction	S4: GD, SGD, feature scaling, and regularization for regression
3	S5: Logistic regression and predictive probabilities	S6: Regression and classification metrics, thresholds, and imbalanced data
4	S7: Decision trees, splitting, impurity, and overfitting	S8: Random forests, bagging, and the bias-variance trade-off
5	S9: Gradient boosting and tabular-data baselines	S10: Cross-validation, hyperparameter search, and experiment tracking
6	S11: Feature transformation, missing data, and preprocessing pipelines	S12: PCA, dimension selection, and information-loss evaluation
7	S13: K-means and nearest neighbors	S14: TF-IDF, retrieval baselines, and search evaluation metrics
8	S15: Error analysis, calibration, and errors across data groups	S16: Defend the ML pipeline and model selection decisions

Lab

Tools & access

Prepared ML environment, datasets, experiment tracking, and a list of suitable contributions.

Evidence you keep

Benchmark and first contribution package.