Towards Causal Representation Learning

Abstract

The two fields of machine learning and graphical causality arose and developed separately. However, there is now cross-pollination and increasing interest in both fields to benefit from the advances of the other. In the present paper, we review fundamental concepts of causal inference and relate them to crucial open problems of machine learning, including transfer and generalization, thereby assaying how causality can contribute to modern machine learning research. This also applies in the opposite direction: we note that most work in causality starts from the premise that the causal variables are given. A central problem for AI and causality is, thus, causal representation learning, the discovery of high-level causal variables from low-level observations. Finally, we delineate some implications of causality for machine learning and propose key research areas at the intersection of both communities.

arXiv commentSpecial Issue of Proceedings of the IEEE - Advances in Machine Learning and Deep Neural Networks

Cited by in corpus (15)

• Towards Out-Of-Distribution Generalization: A Survey
• Learning Disentangled Representations in the Imaging Domain
• Causal inference in drug discovery and development
• Assaying Out-Of-Distribution Generalization in Transfer Learning
• Scene Transformer: A unified architecture for predicting multiple agent trajectories
• Counterfactual Invariance to Spurious Correlations: Why and How to Pass Stress Tests
• Artificial Intelligence in Healthcare: Lost In Translation?
• Direct Advantage Estimation
• Optimization-based Causal Estimation from Heterogenous Environments
• A Topological Perspective on Causal Inference
• Typing assumptions improve identification in causal discovery
• Post-discovery Analysis of Anomalous Subsets
• Using Human-Guided Causal Knowledge for More Generalized Robot Task Planning
• Individual and Collective Autonomous Development
• Understanding the World Through Action