AI Paper Atlas

Research map / Label-efficient and robust learning / Domain adaptation and generalization

Confidence Calibration for Domain Generalization Under Covariate Shift

Yunye Gong, Xiao Lin, Yi Yao, Thomas G. Dietterich, Ajay Divakaran, Melinda Gervasio

ICCV 2021 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Source-free domain adaptation · 31 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
calibrate with temperatures from source domain only instead of leaving uncalibrated (ablation)temperature scaling · Office-HomeECE (%): 8.66 -> 10.45Hurt
add multiple calibration domains (set-level) versus source-only calibrationtemperature scaling · Office-HomeECE (%): 10.45 -> 6.80Improved
add multiple calibration domains (set-level) versus source-only calibrationtemperature scaling · DomainNetECE (%): 16.93 -> 7.83Improved
use cluster-level nearest-neighbor temperature instead of a single set-level temperatureset-level calibration · DomainNetECE (%): 7.83 -> 6.44Improved
use cluster-level nearest-neighbor temperature instead of a single set-level temperatureset-level calibration · Office-HomeECE (%): 6.80 -> 6.98No clear effect
replace NN cluster selection with linear regression of temperature (ablation)cluster-level calibration · DomainNetECE (%): 6.44 -> 9.34Hurt
ensemble logits of set, NN and regression methodscalibration methods · Office-HomeECE (%): 6.80 -> 6.53Improved
use our multi-domain calibration instead of TransCalTransCal · Office-HomeECE (%): 15.39 -> 6.80Improved
use our multi-domain calibration instead of WTSWTS · Office-HomeECE (%): 9.78 -> 6.53Improved
add domain-invariant features to the classifier before calibration (ablation)Ours · StarCraft2ECE (%): 6.30 -> 4.41Improved
change number of clusters (ablation)cluster-level calibration · Office-Home / DomainNetECE (%)No clear effect
train classifier on multiple source domains instead of one (ablation)classifier · Office-Homeaccuracy (%): 51.55 -> 63.35Improved

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