Talks That Caught My Attention
1. Shapeshifter: A Dynamic Graph-Restructuring Algorithm for Improved Inference in Loopy-Structured PGMs
Katerine Marí Sadie, Johan Adam du Preez, Willie Brink
I found this paper very interesting, and Katerine presented it very well. In my opinion, it is probably the best paper of PGM 2026 (on the last day it was selected for an honorable mention for student papers). It performs better than junction trees in practice! I will definitely read it carefully — and I already have an idea for using the proposed approach in my own research.
2. Learning Causal Structure of Time Series using Best Order Score Search
Irene Gema Castillo Mansilla, Urmi Ninad
Another reminder that I should pay closer attention to time-series models and causality in my research. Well done, Irene — for a master’s student, this is impressive work!
3. Counterfactual Fairness under Hidden Confounding
Marc Braun, Jose Peña, Adel Daoud
Interesting and inspiring work by Peña and his team, as always.
Conference Dinner & Private Tour
A lovely conference dinner, followed by a private tour by Gherardo Varando, one of the main organizers of PGM 2026. Thanks for everything, Gherardo, and congratulations on the great work!