Machine learning
Graduate · Data science
Syllabus focus
Topics typically covered
Standard syllabus
Foundations
- Statistical learning framework: risk, ERM, and generalization
- Bias–variance tradeoff and complexity control
- Linear models, kernels, and regularization theory (intro)
- Optimization for ML: GD, SGD, and convexity survey
- Probabilistic models and MLE/MAP intuition
- Representation learning motivation
Core methods
- Trees, boosting, and random forests in depth
- SVMs and kernel methods (course-dependent)
- Neural networks: backprop and architectures survey
- Unsupervised learning: clustering, PCA, embeddings
- Sequence and graph model survey (optional)
- Generative models overview (VAE/GAN/diffusion survey)
Evaluation and theory practice
- Proper scoring rules and calibration
- Cross-validation theory pitfalls
- Multiple testing in model search
- Domain shift and robustness (intro)
- Interpretability methods and limits
- Reproducibility of ML experiments
STEM / applied
Implementation depth
- Implementing algorithms from pseudocode
- Modern libraries (PyTorch/sklearn/jax survey per course)
- Distributed training awareness
- Hyperparameter search at scale
- Benchmarking against strong baselines
- Writing clear experimental sections
Responsible ML
- Dataset documentation and leakage audits
- Fairness interventions survey
- Security: adversarial examples awareness
- Environmental cost of large models (survey)
- Deployment constraints in regulated settings
- Research paper reading and critique
Notes
More theoretical than undergrad Applied ML. Exact elective topics (deep learning emphasis vs classical) follow the offering.