AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Deconfounded Representation Similarity for Comparison of Neural Networks

Tianyu Cui, Yogesh Kumar, Pekka Marttinen, Samuel Kaski

NeurIPS 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Contrastive representation theory · 3 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
Deconfound CKA (CKA -> dCKA)CKA · ImageNet-CIFAR pair detection, average over blocksproportion detected: 0.24, 0.41 -> 0.5, 0.51Improved
Deconfound RSA (RSA -> dRSA)RSA · ImageNet-CIFAR pair detection, average over blocksproportion detected: 0.02, 0.15 -> 0.25, 0.26Improved
Deconfound CKA in deep blocksCKA · ImageNet-CIFAR pair detection, blocks 5-8proportion detected: 0.0, 0.0 -> 0.0, 0.04No clear effect
Deconfound CKA across noise levelsCKA · Cross-domain NN identification (CIFAR-10-C)proportion identified: 0.18 -> 0.29Improved
Deconfound RSA across noise levelsRSA · Cross-domain NN identification (CIFAR-10-C)proportion identified: 0.16 -> 0.23Improved
Deconfound CKA on same-domain inputsCKA · Same-domain NN identificationproportion identified: 0.65 to 0.8Improved
Deconfound RSA on same-domain inputsRSA · Same-domain NN identificationproportion identified: 0.75 to 0.8Improved
Deconfound CKA for transfer learningCKA · DomainNet domain similarity correlationSpearman rho: 0.804 -> 0.895Improved
Deconfound CKA for transfer learningCKA · Multi STS-B domain similarity correlationSpearman rho: 0.358 -> 0.924Improved
Deconfound CKA for OOD accuracy predictionCKA · CIFAR-10-C average over corruption levelsSpearman rho: 0.139 -> 0.145Improved
Deconfound CKA for in-distribution accuracyCKA · CIFAR-10 ID accuracySpearman rho: 0.163 -> 0.167No clear effect
Add higher-order polynomial terms of input similarity to the regression (ablation)deconfounding regression · ResNet-18 CIFAR-10R^2 and BICNo clear effect

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