For the lectures, we take a flipped classroom approach. The overall goal is for you to have worked through the lecture notes and the computer labs after the first 5 weeks. We have video recordings on the topics of the lecture notes to help you learn better. The recordings follow the lecture notes closely but watching the videos does not replace the study of the lecture notes.
All content in the lecture notes and computer labs is examinable unless otherwise noted, and will be assessed in a time-limited test in teaching week 6.
A weekly study guide is provided below. The live activities and the deadlines of the assessed quizzes are summarised in the following table. Please note that the quizzes need to be completed by the set deadlines to count as engagement with the course.
| Teaching week | Dates | Tuesday | Wednesday | Thursday |
|---|---|---|---|---|
| week 1 | Jan 12–14 | Opening–live | Q&A–live | |
| week 2 | Jan 19–21 | Quiz 1 DL | Lab 1–live | Q&A–live |
| week 3 | Jan 26–28 | Quiz 2 DL | Lab 2–live | Q&A–live |
| week 4 | Feb 2–4 | Lab 3–live | Q&A–live | |
| week 5 | Feb 9–13 | Quiz 3 DL | Lab 4–live | Q&A–live |
| flex learning wk | Feb 15–21 | |||
| week 6 | Feb 23–25 | Class test |
The opening and the Q&A live sessions run from 11:10 to 12:00 GMT.
In the Q&A sessions, we will answer the most important questions from piazza first, and are then opening the floor for live questions. To make sure that your questions are considered, please post them by Thursday 8am GMT on piazza.
Week 1: Jan 11–15
Goals this week
- You have signed up to piazza.
- You can access the lecture recordings via learn.
- You have access to the gradescope page of DME via learn.
- You have worked through Chapter 1 of the lecture notes. Main topics: introduction to the data analysis process, simple descriptions and preprocessing of data.
Supporting video recordings (access via learn)
- 01 – Opening, numerical data description
- 02 – Numerical data description (cont)
- 03 – Multivariate measures and data preprocessing (please ignore everything attendance related at the start of the video)
Week 2: Jan 18–22
Goals this week
- You have worked through Sections 2.1 to 2.3 (including) of the lecture notes. Main topics: different approaches to principal component analysis.
- You have worked through computer lab 1 on simple data descriptions and preprocessing.
- You have completed Quiz 1 by Tue 19 Jan, 11:59 PM GMT.
Engagement checkpoint: Quiz 1 is on Chapter 1 of the lecture notes. The deadline for completing the quiz is Tue 19 Jan, 11:59 PM GMT. The quiz is available on gradescope (access via learn) and you can complete it any time before the deadline.
Please note that you have 1 attempt at the quiz and you will need to complete it within 90 minutes once started. The quiz is short. You can use the extra time to familiarise yourself with the system.
For a general introduction to gradescope, see here, in particular the video on “submitting an online assignment”, which is the system mainly used in this course.
Supporting video recordings (access via learn)
- 04 – Principal component analysis: sequential variance maximisation
- 06 – Principal component analysis: equivalent views
- 07 – Principal component analysis: equivalent views (cont), probabilistic PCA: until start of probabilistic PCA (~min 20)
Recording 5 in the collection is a guest lecture by Chris Williams (non-examinable).
Week 3: Jan 25–29
Main goals this week
- You have worked through Sections 2.4 and 3.1 of the lecture notes. Main topics: probabilistic PCA, linear dimensionality reduction.
- You have completed computer lab 2 on principal component analysis.
- You have completed Quiz 2 by Tue 26 Jan, 11:59 PM GMT.
Engagement checkpoint: Quiz 2 is on Sections 2.1 to 2.3 (including) of the lecture notes and computer lab 1. The deadline for completing the quiz is Tue 26 Jan, 11:59 PM GMT.
The quiz is available on gradescope (access via learn) and you can complete it any time before the deadline.
Please note that you have 1 attempt at the quiz and you will need to complete it within 90 minutes once started.
Supporting video recordings (access via learn)
- 07 – Principal component analysis: equivalent views (cont), probabilistic PCA: rest
- 08 – Probabilistic PCA
- 09 – Linear dimensionality reduction (please ignore everything tophat/attendance related in the video)
Week 4: Feb 1–5
Main goals this week
- You have worked through the rest of Chapter 3 of the lecture notes. Main topics: dimensionality reduction by kernel PCA, multidimensional scaling, Isomap, UMAP.
- You have completed computer lab 3 on dimensionality reduction.
Supporting video recordings (access via learn)
- 10 – Nonlinear dimensionality reduction
- 11 – Nonlinear dimensionality reduction (cont)
Week 5: Feb 8–12
Main goals this week
- You have worked through Chapter 4 of the lecture notes. Main topics: performance evaluation in predictive modelling, hyperparameter/model selection
- You have completed the associated computer lab 4.
Engagement checkpoint: Quiz 3 covers Chapter 3 of the lecture notes and computer labs 2 & 3. The deadline for completing the quiz is Tue 9 Feb, 11:59 PM GMT.
The quiz is available on gradescope (access via learn) and you can complete it any time before the deadline.
Please note that you have 1 attempt at the quiz and you will need to complete it within 90 minutes once started.
Supporting video recordings (access via learn):
- 12 – Prediction loss and training loss
- 13 – Generalisation performance
- 14 – Hyperparameter selection and loss functions
(Please ignore everything tophat/attendance related in the videos.)
Week 6: Class-test on Thursday Feb 25
The class test will be held on gradescope, covering the lectures notes and the computer labs.
-
You will be given
80 minutes90 minutes to complete the test. The test is designed to only take 60 minutes. The extra 30 minutes are a buffer to deal with possible technical issues and to have enough time to upload files to the system. -
The test will open at 1am GMT and close at 11pm GMT on Thu Feb 25. Like in the quizzes, you will have one attempt and you will need to complete the test within 90 minutes once started.
-
A practice test on gradescope familiarises you with all the gradescope-features that you will need in the test, including uploading of pictures or typing of latex. The practice test closes Feb 22 at 11pm GMT. Please do it before.
The level of math will be similar to the exams in the previous years, which you can find here (search for “data mining and exploration”, exams prior to 2013/14 are not relevant). Unlike in some exams in previous years, you will have to answer all questions (and not have to select between e.g. question 2 and 3).
You will not be asked to code during the test. But you are expected to be able to read code.
Supporting video recordings (access via learn):
- 15 – Lecture recap
- 16 – Example exam with solutions, and info on expected detail of the answers