Adrián Javaloy "Causal Generative Models: from Theory to Practice"

Machine Learning Coffee seminars are weekly seminars organised by the Finnish Center for Artificial Intelligence FCAI and are held jointly by Aalto University and the University of Helsinki. The seminars aim to gather people from different fields of science with interest in machine learning.
MLCS

Please join us on Monday 23.11.2026 in Exactum, room C220, from 2pm.

Speaker: Adrián Javaloy (University of Bath, UK)

When: Monday November 23 at 14

Where: C220 (Exactum) & Zoom (https://helsinki.zoom.us/j/66684518817?pwd=Naqv9MzCoMxq1ZCS92Z4UTAGenVPeA.1)

Title: Causal generative models: from theory to practice

Abstract: Generative AI excels at capturing complex data distributions, but it typically says little or nothing about what happens when we interact with the world. In this talk, I will explore how we can build (deep) generative models that are causally grounded, enabling their use in causal inference tasks, i.e., in answering interventional and counterfactual queries. We first revisit Causal Normalizing Flows, a family of models that can learn the structural causal mechanisms generating our observations, can be intervened upon, and yield identifiability guarantees thanks to principled assumptions. Then, I will introduce DeCaFlow, which expands this model family to allow for hidden confounders while still enabling principled causal inference through proxy information. Finally, I will end the talk introducing ongoing work that explores these same ideas in settings where only discrete data is available.

Speaker: Adrián recently started working as a Lecturer in Artificial Intelligence at the University of Bath. He is also a member of the Centre for AI at Bath and a member of ELLIS. Before, he worked as a postdoctoral researcher at the University of Edinburgh in the april lab with Dr. Antonio Vergari. He obtained his PhD in Computer Science from Saarland University in 2025, advised by Prof. Isabel Valera, after working with her at the Empirical Inference department of Prof. Bernhard Schölkopf at the Max Planck Institute for Intelligent Systems in Tübingen. Adrián's research is human-driven and focuses on developing machine learning methods that are reliable, principled, and efficient. To this end, his works has involved probabilistic and tractable models, causality, and Riemannian optimisation, but he actually works in any topic that piques his interest, go chat with him!

  • Updated:
  • Published:
Share
URL copied!