Center for Advanced Intelligence Project

This talk will be held only on-line by Zoom.
Language: English

Date & Time: July 09, 2025: 17:30 – 18:30 (JST)

Speaker: Antonio Vergari , University of Edinburgh, UK

Title: From tensor factorizations to circuits (and back)

Abstract: Tensor factorizations are a staple in many data science applications and probabilistic ML and have been extended to hierarchical models via tensor networks. In this talk, I will connect them with another prominent family of probabilistic models: circuits. On the surface, tensor factorizations and circuits look very different and have established two different communities that rarely talk to one another. I will show that these are two faces of the same coin and these communities are better to talk to one another! Under this interpretation, popular factorization methods for probabilistic modelling can be reduced into just two simple characteristics: their circuit structure and how they are parameterized. Then, by navigating different combinations of structures and parameterizations, we can build novel deep factorizations that are hard to express in the usual language of tensor networks, we can easily treat old factorizations as generative models and we can clearly theoretically characterize their expressiveness thanks to the structural properties in the language of circuits.

Bio:
Antonio Vergari is a Reader (Associate Professor) in Machine Learning and a member of the ELLIS Unit at the University of Edinburgh. His research focuses on efficient and reliable machine learning in the wild, tractable probabilistic modeling and combining learning with complex reasoning. He is interested in unifying probabilistic reasoning. Recently, he has been awarded an ERC Starting Grant called “UNREAL – a Unified REAsoning Layer for Trustworthy ML”. Previously he has been a postdoc in the StarAI Lab lead by Guy Van den Broeck at UCLA. And before that he did a postdoc at the Max Planck Institute for Intelligent Systems in Tuebingen, Germany supervised by Isabel Valera. He obtained a PhD in Computer Science and Mathematics at the University of Bari, Italy. He published several conference and journal papers in top-tier AI and ML venues such as NeurIPS, ICML, UAI, ICLR, AAAI, ECML-PKDD and more, some of which have been awarded oral and spotlight presentations. He frequently engages with the tractable probabilistic modeling and the deep generative models communities by organizing a series of events: the Tractable Probabilistic Modeling Workshop (ICML2019, UAI2021, UAI2022, UAI2023), the Tractable PRobabilistic Inference MEeting (T-PRIME) at NeurIPS 2019 and presented a series of tutorials on complex probabilistic reasoning and models at UAI 2019, AAAI 2020, ECAI 2020, IJCAI 2021 and NeurIPS 2022. He organized a Dagstuhl Seminar on “Recent Advancements in Tractable Probabilistic Inference” with Priyank Jaini, Kristian Kersting and Max Welling.