Kernel Methods For Causal Inference
summary
Proposes a practical method that turns long‑term causal questions into three matrix computations: use kernel ridge regressions to learn a surrogate regression, estimate a conditional mean embedding, then average to get dose responses or treatment effects. Works with discrete or continuous treatments and high‑dimensional surrogates, gives closed-form solutions via RKHS tensor products, uses a sequential mean embedding, and proves uniform consistency for dose responses and root‑n, semiparametric efficiency for treatment effects; worst‑case nonparametric scaling is about N^-1/6 rate.
AI-written summary · about
chapters (9)
- 0:00Intro
- 0:33Motivation: Project STAR
- 4:14Main idea
- 7:22Model: Data setting
- 9:43Model: Identification
- 14:04Algorithm: RKHS
- 17:28Algorithm: Tuning
- 18:00Theory: Assumptions
- 21:58Theory: Uniform consistency
related talks
11mCausal Inference For Socio-Economic & Engineering Systems
52mCausal Matrix Estimation
01h 22mGeneralization IV
47mCausal Matrix Completion: Applications to Offline Causal Reinforcement Learning
31mA Gentle Introduction To Causal Inference
53mAdversarial Machine Learning and Instrumental Variables for Flexible Causal Modeling
01h 01mThe Mathematics of Causal Inference, with Reflections on Machine Learning and the Logic of Science
33mThe Power and Limitations of Kernel Learning