AI Paper Atlas

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

Boosted CVaR Classification

Runtian Zhai, Chen Dan, Arun Suggala, J Zico Kolter, Pradeep Kumar Ravikumar

NeurIPS 2021 · Learning theory and algorithms › Prediction and decision losses · approach: Losses for decision-making · 0 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 alpha-AdaLPBoost ensemble instead of single ERM model at small alphaalpha-AdaLPBoost · COMPAS, CelebA, CIFAR-10, CIFAR-100test alpha-CVaR zero-one lossImproved
use alpha-AdaLPBoost instead of ERM at large alphaalpha-AdaLPBoost · all four datasetstest alpha-CVaR zero-one lossNo clear effect
choose model weights with LPBoost instead of uniform averaging (AdaBoost + Average) (ablation)alpha-AdaLPBoost · CIFAR-10, CIFAR-100test average lossImproved
choose model weights with LPBoost instead of uniform averaging (ablation)alpha-AdaLPBoost · large alphaalpha-CVaR zero-one lossImproved
replace Regularized LPBoost (beta=100) with alpha-AdaLPBoost (ablation)alpha-AdaLPBoost · CIFAR-10alpha-CVaR zero-one lossNo clear effect
increase number of base models T beyond 30 (ablation)alpha-AdaLPBoost · CIFAR-10, CIFAR-100test alpha-CVaR zero-one lossNo clear effect
train ensemble with T=30 base modelsAdaBoost · training costtraining time relative to ERM: 3-6 times the training time of ERMHurt

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