AI Paper Atlas

Research map / Learning theory and algorithms / Sample complexity

Dropout Training is Distributionally Robust Optimal

José Blanchet, Yang Kang, José Luis Montiel Olea, Viet Anh Nguyen, Xuhui Zhang

JMLR 2023 · Learning theory and algorithms › Sample complexity · approach: Optimal mean estimation and subsampling

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
choose dropout probability delta=c/sqrt(n) with c=z_{1-alpha}*sigma/mu instead of 10-fold cross-validationdropout training in linear regression · Coverage of population loss, n=1000coverage frequency: 0.52 -> 0.88 (alpha=0.1)Improved
choose dropout probability delta=c/sqrt(n) instead of plain OLSlinear regression · Coverage of population loss, n=10000coverage frequency: 0.47 -> 0.90 (alpha=0.1)Improved
use unbiased multi-level Monte Carlo estimator instead of standard SGD for dropout trainingdropout training in high-dimensional linear regression (d=100, n=50) · parameter divergence from beta*_n, 60 s wall-clockl2 divergence: outperforms SGD once the number of parallel iterations exceeds ~1000Improved
increase number of parallelizations L beyond about 2000 for multi-level Monte Carlo (ablation)unbiased multi-level Monte Carlo dropout · parameter divergence from beta*_ndivergence: saturation after ~2000 parallelizationsNo clear effect

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