Generative models and uncertainty quantification lie at the heart of Bayesian modelling and inference. At this small meeting, we discuss recent developments within the field. The workshop is deliberately kept small to ensure the discussions remain honest, lively and interesting. Attendance is, thus, primarily by invitation, but one can apply to join (see below).
A safe environment is a prerequisite for engaging discussions. Participants are, thus, expected to abide by the DDSA Code of Conduct.
Speakers
Program
The program below is tentative and subject to change.
| Wednesday (Sep 16) | Thursday (Sep 17) | ||
|---|---|---|---|
| 08:30-08:55 | Registration and coffee |
08:30-09:00 | Coffee |
| 08:55-09:00 | Opening remarks Wouter Boomsma |
||
Session 1: Diffusion models Chairs: Wouter Boomsma and Jes Frellsen |
Session 3: Neuro-symbolic AI Chair: Wouter Boomsma |
||
| 09:00-10:00 | Alexander Tong On Couplings for Flow Models |
09:00-10:00 | Antonio Vergari Neuro-symbolic AI from first principles |
| 10:00-10:30 | Coffee |
10:00-10:30 | Coffee |
Session 4: Controllable and efficient generation Chair: Nico Krämer |
|||
| 10:30-11:30 | Molei Tao Understanding and enhancing diffusion model: a quantification of its creativity, and a provably scalable test-time scaling method |
10:30-11:30 | Damien Garreau Towards Understanding Steering Strength |
| 11:30-12:30 | Yee Whye Teh Exact Posterior Score Estimation for Solving Linear Inverse Problems |
11:30-12:30 | Sungsoo Ahn Amortizing molecular simulations with density fields |
| 12:30-13:30 | Lunch |
12:30-13:30 | Lunch |
Session 2: Does GenAI need UQ? Chair: Ole Winther |
Session 5: AI for scientific discovery Chair: Jes Frellsen |
||
| 13:30-14:30 | Eyke Hüllermeier The Aleatoric-Epistemic Dichotomy of Uncertainty Quantification in Machine Learning and GenAI: Challenges, Doubts, and Opportunities |
13:30-14:30 | Kevin Murphy Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models |
| 14:30-15:00 | Coffee |
14:30-15:00 | Coffee |
Session 6: Adaptive learning Chair: Søren Hauberg |
|||
| 15:00-16:00 | Henry Moss Ask not what genAI can do for Gaussian processes —ask what Gaussian processes can do for GenAI |
15:00-16:00 | Emtiyaz Khan Adaptive Intelligence |
| 16:00-17:00 | Maike Osborne AI governance needs Bayes? |
16:00-17:00 | Siddharth Swaroop Knowledge adaptation using a Bayesian lens |
| 17:00-18:30 | Poster session with reception |
17:00 | Parting remarks and goodbye Søren Hauberg |
| 19:00-21:30 | Dinner at Madenitaly |
||
Registration
If you would like to attend this workshop but have yet to receive an invitation, then please write an email to Søren Hauberg and include a link to your website or Google Scholar profile. Note that seating is limited, so we cannot guarantee a ticket.
Venue
Carlsberg Akademi
Gamle Carlsberg Vej 15
1799 Copenhagen V
Carlsberg Akademi is less than 10 minutes walk from Carlsberg S-train Station.
Dinner
Restaurant Madenitaly
Kødbyen H15
Slagtehusgade 1
1700 Copenhagen V
Italian Vegan Restaurant with a twist.
Organisation
The workshop is jointly organised by:
Funding
The workshop is funded by the Danish Data Science Academy, the Carlsberg Foundation, and the Center for Basic Machine Learning Research in Life Science.