Dear students, 

Luis, Max and I have updated our lecture plan,

These items have changed: 

• We will have an optional Q&A session on Perceptron and MLPs on 21 Dec 2020.
• The mid-term exam will be published 4 Jan 2020, 10:00 and required to be submitted 13:00.  See below for more details. 
• All lecture Q&As of the second half are shifted by one week (see below)
• The final exam will be published 27 Feb 2020, 10:00 and required to be submitted 13:00. 

New Q&A dates (videos will be made available in the week before, assignments are due at least one week later):

02 Nov 2020 - Introduction, notation, k-nearest neighbors
09 Nov 2020 - Clustering (kMeans, DBSCAN)
16 Nov 2020 - Linear and logistic regression
23 Nov 2020 - Model validation
30 Nov 2020 - The covariance matrix, PCA
07 Dec 2020 - Bagging, decision trees, random forests
14 Dec 2020 - Boosting (AdaBoost), Viola-Jones
21 Dec 2020* - Perceptron, multi-layer perceptron
11 Jan 2021 - Gradient Descent, Backprop, Optimizers (SGD, Adam, RProp)
18 Jan 2021 - ConvNets
25 Jan 2021 - Unsupervised representation learning I  (VAEs, Glow)
01 Feb 2021 - Unsupervised representation learning II (GANs)
08 Feb 2021 - RNNs
15 Feb 2021 - Reinforcement Learning, Policy Gradients
22 Feb 2021 - Attribution, Adversarial Examples (BONUS)

Mid-term and final exams

After a lengthy discussion we agreed to:

• conduct exams in a 3 hours window (as opposed to a day or more)
• you download the questions at 10:00 and submit your answers to the whiteboard not later than 13:00 
• You will find a mix of multiple choice and full-text questions

Why don't we get more time?

The reasons we decided for this solution are

• 3 hours are enough time for something that we usually do in 45 min
• giving you more time creates potential for unfairness and requires the questions to be harder to copy/paste from the web

So, in other words: exams will be shorter and easier. 

Take care, 

Max, Luis and Tim