AI Paper Atlas

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

Wasserstein Logistic Regression with Mixed Features

Aras Selvi, Mohammad Reza Belbasi, Martin B Haugh, Wolfram Wiesemann

NeurIPS 2022 · Learning theory and algorithms › Prediction and decision losses · approach: Distributionally robust optimization · 1 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 Wasserstein DR mixed-feature logistic regression (kappa = m) instead of unregularized LRlogistic regression · UCI categorical benchmarks (14 datasets)share of instances with lowest error: 21% -> 79%Improved
use DR logistic regression with output label weight kappa = 1 instead of LR (unregularized)logistic regression · UCI categorical benchmarks (14 datasets)share of instances with lowest error: 21% -> 36%Improved
use output label weight kappa = m instead of kappa = 1 in the ground metric (ablation)DR logistic regression · UCI categorical benchmarks (14 datasets)share of instances with lowest error (unregularized): 36% -> 79%Improved
use regularized r-DRO (either kappa) instead of Lasso r-LRLasso logistic regression · UCI categorical benchmarks (14 datasets)share of instances with lowest error (regularized): 14% -> 57%Improved
compare to mass-transportation regularized (MT) baseline kappa = 1 vs mMT regularization · UCI categorical benchmarks (14 datasets)share of instances with lowest error: 42% vs 57%Mixed
compare to Wasserstein profile inference (PI) baselinePI regularization · UCI categorical benchmarks (14 datasets)share of instances with lowest error: 14%Hurt
use DRO (kappa = m) instead of LRlogistic regression · UCI lymphographyclassification error: 21.17% -> 16.72%Improved
use r-DRO (kappa = m) instead of r-LRLasso logistic regression · UCI spectclassification error: 17.79% -> 15.83%Improved
use DRO (kappa = m) instead of LRlogistic regression · UCI audiologyclassification error: 14.18% -> 14.40%No clear effect
use DRO (kappa = m) instead of LRlogistic regression · UCI carclassification error: 4.73% -> 4.73%No clear effect
use DRO (kappa = m) instead of LRlogistic regression · UCI splicesclassification error: 6.80% -> 6.66%Improved
use r-DRO (kappa = m) instead of r-LRLasso logistic regression · UCI house-votes-84classification error: 4.54% -> 5.16%Hurt

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