Recent advances in theoretical machine learning
This is a weekly meeting / reading group of the TFAI group, where we discuss new papers and occasionally have invited talks.
Everyone in TUM (and beyond) is welcome to join. Just come to the venue at the specified time. Please introduce yourself if you are joining for the first time. For best experience, read the paper before the meeting.
Time and place
Wednesdays 10:00-12:00
Most meeting take place only in-person in Room 03.11.058
Occasionally (specified below), online participation is possible via Zoom [link ; Passcode: 221241]
Upcoming meetings
22.10. We will start with an introduction to finite-state automata (DFAs, NFAs, and transducers) and then read Transformers as Transducers.
29.10 There will be short presentations on two papers (moderated by two bachelor students)
- 10:00 Abbe, Sandon. On the universality of deep learning. (We will cover till section 2.2)
- 10:45 Hanin, Zlokapa. Bayesian Interpolation with Deep Linear Networks. (We will cover till page 11)
5.11 no meeting
12.11 TBA
TBA. Paper reading: A mathematical perspective on Transformers
TBA. Tutorial on the paper Computational Complexity of Statistics: New Insights from Low-Degree Polynomials