AI Paper Atlas

Research map / Learning theory and algorithms / Prediction and decision losses

RieszNet and ForestRiesz: Automatic Debiased Machine Learning with Neural Nets and Random Forests

Victor Chernozhukov, Whitney Newey, Vı́ctor M Quintas-Martı́nez, Vasilis Syrgkanis

ICML 2022 · Learning theory and algorithms › Prediction and decision losses · approach: Distributionally robust optimization · 5 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use the doubly robust moment rather than the direct plug-in momentRieszNet · IHDP ATEMAE: 0.123 -> 0.110Improved
Use the doubly robust moment rather than the direct plug-in momentForestRiesz · IHDP ATEMAE: 0.197 -> 0.126Improved
Use RieszNet DR instead of DragonnetIHDP ATE · IHDP ATEMAE: 0.146 ± 0.010 -> 0.110 ± 0.003Improved
Use ForestRiesz DR instead of RF plug-inIHDP ATE · IHDP ATEMAE: 0.389 ± 0.024 -> 0.126 ± 0.004Improved
Add the post-processing TMLE adjustment to DRForestRiesz · BHP complex f with linear and non-linear confoundersRMSE: 0.377 -> 0.327Improved
Add the post-processing TMLE adjustment to DRForestRiesz · BHP complex f with linear and non-linear confounderscoverage: 0.909 -> 0.953Improved
Replace shared multitasking networks with separate regression and Riesz networks (ablation)RieszNet · IHDP ATERMSE: 0.147 -> 0.411Hurt
Freeze shared layers after training on the Riesz loss (ablation)RieszNet · IHDP ATERMSE: 0.147 -> 1.221Hurt
Replace end-to-end TMLE adjustment training with post-processing (ablation)RieszNet · IHDP ATEbias: -0.044 -> -0.088Hurt
Remove cross-fitting (ablation)ForestRiesz DR · BHP complex f with linear and non-linear confounderscoverage: 0.909 -> 0.894Hurt
Remove cross-fitting while retaining multitasking (ablation)ForestRiesz DR · BHP complex f with linear and non-linear confounderscoverage: 0.909 -> 0.884Hurt
Remove multitasking while retaining simple cross-fitting (ablation)ForestRiesz DR · BHP complex f with linear and non-linear confoundersRMSE: 0.377 -> 0.365Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy