Towards Credible and Effective Data-Driven Decision-Making

on YouTube ↗09m

" summary

Describes a concrete method for minimax-optimal policy learning under unobserved confounding: parameterize confounding with a sensitivity model, plug that into a robustified version of inverse propensity weighting, and solve a robust ERM to pick policies that minimize worst-case regret. Gives decision-theoretic guarantees (out-of-sample improvement versus baseline when the uncertainty set is well specified) and statistical rates for the optimistic estimator, plus a case study comparing an observational study to a clinical trial.

AI-written summary · about

" chapters (9)

1. 0:00Introduction
2. 0:38Challenges of causal inference
3. 3:08Unobserved confounders
4. 3:42Recap
5. 4:12Performance Metrics
6. 4:53Robust Machine Learning Recipe
7. 6:46Guarantees
8. 7:41Interleaving
9. 7:57Conclusion

" related talks

• 28mConfounding-Robust Policy Evaluation In Infinite-Horizon Reinforcement LearningAngela Zhou · Algorithmic Aspects of Causal Inference 2022
• 53mOff-policy Policy OptimizationDale Schuurmans · Emerging Challenges in Deep Learning (Simons Institute workshop, 2019)
• 17mCodifying Data Science IntuitionRyan West · MLconf Atlanta 2017
• 40mDecision-Making Under MiscalibrationGal Yona · Algorithmic Aspects of Causal Inference 2022
• 29mPure Exploration ProblemsWouter Koolen · Mathematics of Online Decision Making 2020
• 55mZeroth-Order Methods for Convex-Concave Minmax Problems: Learning from Strategically Generated DataChinmay Maheshwari · Adversarial Approaches in Machine Learning 2022
• 41mConfident Off-policy Evaluation and Selection through Self-Normalized Importance WeightingIlja Kuzborskij · Reinforcement Learning from Batch Data and Simulation
• 41mKeynote: Personalization, Bandits, and Causal inferenceMaurits Kaptein · PyData Eindhoven 2019