atlas

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

Agree to Disagree: Demystifying Homogeneous Deep Ensembles through Distributional Equivalence

ICLR 2025 · Learning theory and algorithms › Prediction and decision losses · approach: Distributionally robust optimization

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Analyzes why homogeneous deep ensembles improve performance by framing ensemble members through distributional equivalence, rather than relying only on Jensen gap explanations.

Problem

The origin of performance gains from deep ensembles is poorly understood; existing studies attribute them to Jensen gaps of the ensemble loss.

Approach

Proposes a distributional-equivalence view of homogeneous deep ensembles to explain their improvement, going beyond the Jensen-gap explanation prevalent in prior work.

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