Machine Learning and Causality

The research focus of the Machine Learning (ML) and Causality groups at Columbia Engineering is on the foundations of learning, decision-making, explanation, and generalization and their applications throughout the sciences and society.

The increasing use of ML to make decisions in a variety of human-facing domains has highlighted the concerns of robustness, interpretability, and explainability in ML. At Columbia Engineering, we consider understanding Causality, and its interplay with Machine Learning, as central to addressing some of these key shortcomings of the current state of AI. Our research in ML and Causal Inference focuses on statistical and theoretical foundations, along with new algorithm design and interdisciplinary applications. Specific areas of foundational research in ML include statistical and computational learning theory, online machine learning (online learning, multi-armed bandits, and reinforcement learning), probabilistic machine learning, meta-learning, and representational learning. The research in ML is complemented and enriched by our work in causality, where some research directions include foundational aspects of causal inference, causal reinforcement learning, fairness analysis, and generalizability.

Several current collaborative and interdisciplinary projects aim to transfer the advances in these techniques for real impact in a variety of application domains including computer vision, multimedia knowledge extraction, computational biology, climate science, finance, and robotics.