About Me
Welcome to my research website. I am David Rohde, a researcher specializing in Bayesian inference, causality, and recommender systems. My work focuses on developing machine learning algorithms to solve real-world problems.
More recently my research has focused on system wide optimization of recommender systems or interactive systems. In particular considering:
- Differences between academic conventions and in production recommender systems.
- Incorporating user feedback or bandit feedback (click signals and A/B testing) with other signals (content and collaborative filtering)
- Relevance of causal theory, off policy learning, reinforcement learning to real systems
Quite a lot of this work resulted in commentary in position papers, presentations or podcast discussions. I am currently working on a textbook, that will be a more systematic treatment of reward optimizing recommender systems.
I also have a number of scientific papers, usually with colleagues at Criteo, including two excellent PhD students that I had the good fortune to co-supervise.
I am an enthusiastic advocate of the idea that causal inference is inference.
Papers
Talks
- Why the Shooting in the Dark Method Dominates Practice
- Causal Inference with Bayes rule with Finnian Lattimore and David Rohde
- AI Conversations Podcast (9 Videos)
- RecSys 2020 Tutorial: Bayesian Value Based Recommendation
- David Rohde - Causal Inference is Inference – A beautifully simple idea that not everyone accepts
- Panel: Does causality mean we need to go beyond Bayesian decision theory? (audio only)
- A Gentle Introduction to Recommendation as Counterfactual Policy Learning