Turning Challenges into Opportunities: How Distribution Shifts Enhance Identifiability in Causal Representation Learning

Yuhang Liu, Zhen Zhang, Dong Gong, Biwei Huang, Mingming Gong, Anton van den Hengel, Kun Zhang, Javen Qinfeng Shi

Keywords: Causal Representation Learning, Latent Causal Models, Identifiability, Distribution Shifts

Abstract: Causal representation learning seeks to uncover latent causal variables and their relationships from observed, unstructured data, a task complicated by identifiability challenges. While distribution shifts, viewed as natural interventions on latent causal variables, often present difficulties in traditional machine learning tasks, they also create valuable opportunities for identifiability by introducing variability in latent variables. In this paper, we study a non-parametric condition characterizing the types of distribution shifts that contribute to identifiability within the context of latent additive noise models. We also present partial identifiability results when only a portion of distribution shifts meets the condition. Furthermore, we extend our findings to latent post-nonlinear causal models. Building on our theoretical results, we propose a practical algorithm facilitating the acquisition of reliable latent causal representations. Our algorithm, guided by our underlying theory, has demonstrated outstanding performance across a diverse range of synthetic and real-world datasets. The empirical observations closely align with the theoretical findings, affirming the robustness and effectiveness of our proposed approach.

Primary Area: causal reasoning

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 5594