skip to primary navigation skip to content

Data Science

Lecture notes

Abridged notes as printed — examinable material only
Extended notes with extra material on non-examinable material such as neural networks

If you spot a mistake in the printed notes, let me know.

Announcements and Q&A

— Moodle

Lecture schedule

This is the planned lecture schedule. It will be updated as and when actual lectures deviate from schedule. Material marked * is non-examinable. Slides are uploaded the night before a lecture, and re-uploaded after the lecture with annotations made during the lecture.

Prerequisites
	Example sheet 0 and solutions
§1–§4. Learning with probability models
Lecture 1 [slides]	1. Learning with probability models (4:08) 1.1 Specifying probability models (15:20)
Lecture 2 [slides]	1.2 Standard random variables (3:21) 1.3 Maximum likelihood estimation (17:35) 1.4 Numerical optimization (8:01)
Lecture 3 [slides]	1.5 Likelihood notation (10:00) 1.6 Generative models (8:14) 1.7 Supervised learning (14:18) 3.1, 3.2 Prediction accuracy versus probability modelling (* non-examinable)
Lecture 4 [slides]	Mock exam question 1 and walkthrough (23:36) 3.3 Neural networks (* non-examinable)
Lecture 5 [slides]	2.1 Linear modelling (13:27) 2.2 Feature design (19:39) 2.3 Diagnosing a linear model (5:29)
Lecture 6 [slides]	2.5 The geometry of linear models (12:07) 2.6 Interpreting parameters (20:03)
Lecture 7 [slides]	2.4 Probabilistic linear modelling (9:45)
	Example sheet 1 OPTIONAL ex1 practical exercises [ex1.ipynb] (for supervisions) OPTIONAL PyTorch introduction and challenge OPTIONAL climate dataset challenge climate.ipynb Code snippets: fitting.ipynb, lm.ipynb Datasets investigated: climate.ipynb, stop-and-search.ipynb
§5, §6, §8. Bayesian inference and Monte Carlo
Lecture 8 [slides]	5.1 Bayes's rule for random variables (17:27) 5.2 Bayes's rule calculations (15:37)
Lecture 9 [slides]	6.1 Monte Carlo integration (13:21) Discussion of climate dataset challenge (* non-examinable) 4.1 Measuring model fit (* non-examinable)
Lecture 10 [slides]	6.2 Bayes's rule via computation (19:25) 8.3 Finding the posterior (24:45)
Lecture 11 [slides]	8.1, 8.2 Bayesianism (16:54) 8.4 Bayesian readouts (6:28)
video only	Mock exam question 2 and walkthrough (29:35)
	Example sheet 2 OPTIONAL ex2 practical exercises [ex2.ipynb] (for supervisions) OPTIONAL Climate confidence challenge Code snippets: bayes.ipynb
§7, §9, §10. Frequentist methods: hypothesis testing and empirical evaluation
Lecture 11 ctd [slides]	5.3 Deriving the likelihood (9:17)
Lecture 12 [slides]	7.1–7.2 Empirical cdf (15:10) 7.3 The empirical distribution (6:06) 9. Frequentism (3:58) 9.1, 9.2 Resampling / confidence intervals (13:03) freq.ipynb 9.6 Non-parametric resampling (7:00)
Lecture 13 [slides]	9.3 Hypothesis testing (23:37)
Lecture 14 [slides]	9.3 Hypothesis testing (continued) 10. Holdout evaluation and the challenge of induction (* non-examinable)
video only	Mock exam question 3 and walkthrough (18:20)
	Example sheet 3 OPTIONAL ex3 practical exercises [ex3.ipynb] (for supervisions) OPTIONAL Climate confidence challenge
§11+§12. Markov chains and other sequence models
Lecture 14 ctd [slides]	12.2–3 Sequence models with history: RNNs, HMMs, and Transformers (9:57)
Lecture 15 [slides]	12.1 Learning a Markov chain (4:56) 11, 11.1 Markov chain models (12:09) 11.2–3 Calculations with Markov chains (11:37)
Lecture 16 [slides]	Markov chain behaviour (3:22) 11.4 Stationarity, 11.5* Average behaviour (12:38)
	Example sheet 4 OPTIONAL ex4 practical exercises [ex4.ipynb] (for supervisions) OPTIONAL Stoat-finding challenge and Moodle submission (deadline noon 2023-12-24)