About this Event
Kirkbride Lecture Hall, University of Delaware, Newark, DE 19716, USA
http://cis.udel.eduCausal Representation Learning
ABSTRACT
Machine learning has seen great success in learning empirically useful features from high-dimensional data. However, standard approaches often learn features that do not reliably correspond to the underlying mechanisms of interest—such as stable factors of variation or physically meaningful concepts—limiting interpretability and transfer across settings. Causal representation learning (CRL) addresses these challenges by recovering true latent variables that are identifiable (i.e., unique and well-specified) and that support disentangled and causal structure, along with the relationships governing their interactions. In this talk, I will introduce CRL as an emerging field that combines the strengths of machine learning and causal modeling, and discuss its core goals and challenges. I will show how multi-environment data enable rigorous identifiability guarantees and lead to scalable algorithms across a wide range of settings. A central theme is the connection between CRL and score functions, which bridges CRL with modern generative modeling and yields practical estimators in high-dimensional regimes. Finally, I will demonstrate these ideas in practice with a robotics case study.
BIOGRAPHY
Burak Varıcı is an Assistant Professor in the Department of Electrical and Computer Engineering at the University of Delaware. Before joining UD in 2026, he was a postdoctoral researcher at Carnegie Mellon University and received his Ph.D. in Electrical Engineering from Rensselaer Polytechnic Institute in 2024. His doctoral research was supported by IBM AI Horizons PhD Fellowship and has been awarded the 2025 IEEE Signal Processing Society Best Ph.D. Dissertation Award. His research lies at the intersection of machine learning and causality, focusing on the statistical and algorithmic foundations for learning reliable representations of complex data. Specifically, his work spans causal representation learning, self-supervised learning, causal discovery, and intervention design.
0 people are interested in this event
User Activity
No recent activity