## Reading Course: Theory of Machine Learning

### â‡’ Part of MeTh -- Methods in Theoretical Computer Science

`WARNING: This is a proof-based theory course.`

### Content

We will read the textbook

Chapters 2-8 will be covered contiguously, after that we will horsey around.

The course is proof-based, meaning all of the content will consist of:

- mathematically rigorous definitions of machine learning concepts;
- mathematically proving statements about them.

Generally, the level of abstraction is higher than in practice-oriented machine learning courses. For the homework assignments and practice sessions, students will prove stuff themselves. Moreover, since it is a reading course, students will read the textbook themselves, basically, and in class we only discuss questions which have come up. This differs from "frontally" taught courses where the instructor explains (almost) everything.

`WARNING! Python: 0%. Java: 0%. C/C++: 0%. Proofs: 100%.`

### Prerequisites

You should be fine if:

- you were successful in the usual undergraduate math courses: linear algebra, calculus, probability, discrete math;
- you are comfortable reading, understanding and producing proofs.

The textbook is quite gentle on the background. But make no mistake: it's theory.

### Organization

Both reading & solving is to be done individually (or in groups) at home. In class, we will only discuss questions and solutions.

As usual, please ignore the text on Ã•IS.

Meeting times: text session `Tue 12-14`

; problem session `Fri 12-14`

Classes will take place in Paabel (room to be announced).

Contact: Dirk Oliver Theis `dotheis`

(at) `ut`

(dot) `ee`

.