Matches in SemOpenAlex for { <https://semopenalex.org/work/W4311553386> ?p ?o ?g. }
Showing items 1 to 75 of
75
with 100 items per page.
- W4311553386 abstract "Spurious correlations in training data often lead to robustness issues since models learn to use them as shortcuts. For example, when predicting whether an object is a cow, a model might learn to rely on its green background, so it would do poorly on a cow on a sandy background. A standard dataset for measuring state-of-the-art on methods mitigating this problem is Waterbirds. The best method (Group Distributionally Robust Optimization - GroupDRO) currently achieves 89% worst group accuracy and standard training from scratch on raw images only gets 72%. GroupDRO requires training a model in an end-to-end manner with subgroup labels. In this paper, we show that we can achieve up to 90% accuracy without using any sub-group information in the training set by simply using embeddings from a large pre-trained vision model extractor and training a linear classifier on top of it. With experiments on a wide range of pre-trained models and pre-training datasets, we show that the capacity of the pre-training model and the size of the pre-training dataset matters. Our experiments reveal that high capacity vision transformers perform better compared to high capacity convolutional neural networks, and larger pre-training dataset leads to better worst-group accuracy on the spurious correlation dataset." @default.
- W4311553386 created "2022-12-27" @default.
- W4311553386 creator A5021932110 @default.
- W4311553386 creator A5030284058 @default.
- W4311553386 creator A5037989560 @default.
- W4311553386 creator A5061227377 @default.
- W4311553386 creator A5075765401 @default.
- W4311553386 creator A5081669439 @default.
- W4311553386 creator A5087624455 @default.
- W4311553386 date "2022-12-12" @default.
- W4311553386 modified "2023-09-27" @default.
- W4311553386 title "You Only Need a Good Embeddings Extractor to Fix Spurious Correlations" @default.
- W4311553386 doi "https://doi.org/10.48550/arxiv.2212.06254" @default.
- W4311553386 hasPublicationYear "2022" @default.
- W4311553386 type Work @default.
- W4311553386 citedByCount "0" @default.
- W4311553386 crossrefType "posted-content" @default.
- W4311553386 hasAuthorship W4311553386A5021932110 @default.
- W4311553386 hasAuthorship W4311553386A5030284058 @default.
- W4311553386 hasAuthorship W4311553386A5037989560 @default.
- W4311553386 hasAuthorship W4311553386A5061227377 @default.
- W4311553386 hasAuthorship W4311553386A5075765401 @default.
- W4311553386 hasAuthorship W4311553386A5081669439 @default.
- W4311553386 hasAuthorship W4311553386A5087624455 @default.
- W4311553386 hasBestOaLocation W43115533861 @default.
- W4311553386 hasConcept C104317684 @default.
- W4311553386 hasConcept C117978034 @default.
- W4311553386 hasConcept C119857082 @default.
- W4311553386 hasConcept C124101348 @default.
- W4311553386 hasConcept C127413603 @default.
- W4311553386 hasConcept C153180895 @default.
- W4311553386 hasConcept C154945302 @default.
- W4311553386 hasConcept C185592680 @default.
- W4311553386 hasConcept C21880701 @default.
- W4311553386 hasConcept C41008148 @default.
- W4311553386 hasConcept C51632099 @default.
- W4311553386 hasConcept C55493867 @default.
- W4311553386 hasConcept C63479239 @default.
- W4311553386 hasConcept C81363708 @default.
- W4311553386 hasConcept C8642999 @default.
- W4311553386 hasConcept C95623464 @default.
- W4311553386 hasConcept C97256817 @default.
- W4311553386 hasConceptScore W4311553386C104317684 @default.
- W4311553386 hasConceptScore W4311553386C117978034 @default.
- W4311553386 hasConceptScore W4311553386C119857082 @default.
- W4311553386 hasConceptScore W4311553386C124101348 @default.
- W4311553386 hasConceptScore W4311553386C127413603 @default.
- W4311553386 hasConceptScore W4311553386C153180895 @default.
- W4311553386 hasConceptScore W4311553386C154945302 @default.
- W4311553386 hasConceptScore W4311553386C185592680 @default.
- W4311553386 hasConceptScore W4311553386C21880701 @default.
- W4311553386 hasConceptScore W4311553386C41008148 @default.
- W4311553386 hasConceptScore W4311553386C51632099 @default.
- W4311553386 hasConceptScore W4311553386C55493867 @default.
- W4311553386 hasConceptScore W4311553386C63479239 @default.
- W4311553386 hasConceptScore W4311553386C81363708 @default.
- W4311553386 hasConceptScore W4311553386C8642999 @default.
- W4311553386 hasConceptScore W4311553386C95623464 @default.
- W4311553386 hasConceptScore W4311553386C97256817 @default.
- W4311553386 hasLocation W43115533861 @default.
- W4311553386 hasOpenAccess W4311553386 @default.
- W4311553386 hasPrimaryLocation W43115533861 @default.
- W4311553386 hasRelatedWork W2909857627 @default.
- W4311553386 hasRelatedWork W2972074553 @default.
- W4311553386 hasRelatedWork W2995914718 @default.
- W4311553386 hasRelatedWork W3043252291 @default.
- W4311553386 hasRelatedWork W3081496756 @default.
- W4311553386 hasRelatedWork W3095523211 @default.
- W4311553386 hasRelatedWork W3193301557 @default.
- W4311553386 hasRelatedWork W4223905158 @default.
- W4311553386 hasRelatedWork W4307883119 @default.
- W4311553386 hasRelatedWork W564581980 @default.
- W4311553386 isParatext "false" @default.
- W4311553386 isRetracted "false" @default.
- W4311553386 workType "article" @default.