AI Paper Atlas

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

Understanding Hessian Alignment for Domain Generalization

Sobhan Hemati, Guojun Zhang, Amir Estiri, Xi Chen

ICCV 2023 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Optimization-based domain generalization · 19 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 Hutchinson Hessian-diagonal alignment instead of Fishr/V-RExColored MNIST MLP · Colored MNISTtest accuracy (%): 69.5 -> 74.0Improved
use Hutchinson alignment under heavy label shiftImbalanced Colored MNIST · Imbalanced Colored MNISTtest accuracy (%): 35.5 -> 47.7Improved
use HGP (Hessian-gradient product) instead of FishrColored MNIST · Colored MNISTtest accuracy (%): 69.5 -> 69.4No clear effect
use HGP on DomainBed instead of ERMResNet-50 DomainBed · DomainBed averageaverage accuracy (%): 66.6 -> 66.9No clear effect
use Hutchinson alignment on DomainBed instead of ERMResNet-50 DomainBed · DomainBed averageaverage accuracy (%): 66.6 -> 68.5Improved
add Hessian alignment on top of gradient alignment (ablation)Hutchinson · PACS (test domain sketch)test accuracy (%): 77.0 -> 81.4Improved
add gradient alignment on top of Hessian alignment (ablation)Hutchinson · PACS (test domain sketch)test accuracy (%): 79.4 -> 81.4Improved
add gradient alignment term (beta) to Hutchinson (ablation)Hutchinson · Colored MNISTtest accuracy (%): 9.8 -> 74.0Improved
add Hessian alignment term (alpha) to Hutchinson (ablation)Hutchinson · Colored MNISTtest accuracy (%): 69.3 -> 74.0Improved
add Hessian-gradient-product term (alpha) to HGP (ablation)HGP · Colored MNISTtest accuracy (%): 69.3 -> 69.4No clear effect
use Hutchinson Hessian alignment instead of baselinesCMNIST training cost · CMNIST training time, 100 stepstraining time: ERM 15.3 -> Hutchinson 62.4Hurt

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