Matches in SemOpenAlex for { <https://semopenalex.org/work/W4386083105> ?p ?o ?g. }
Showing items 1 to 90 of
90
with 100 items per page.
- W4386083105 abstract "Recent advances in robust semi-supervised learning (SSL) typically filter out-of-distribution (OOD) information at the sample level. We argue that an overlooked problem of robust SSL is its corrupted information on semantic level, practically limiting the development of the field. In this paper, we take an initial step to explore and propose a unified framework termed OOD Semantic Pruning (OSP), which aims at pruning OOD semantics out from in-distribution (ID) features. Specifically, (i) we propose an aliasing OOD matching module to pair each ID sample with an OOD sample with semantic overlap. (ii) We design a soft orthogonality regularization, which first transforms each ID feature by suppressing its semantic component that is collinear with paired OOD sample. It then forces the predictions before and after soft orthogonality decomposition to be consistent. Being practically simple, our method shows a strong performance in OOD detection and ID classification on challenging benchmarks. In particular, OSP surpasses the previous state-of-the-art by 13.7% on accuracy for ID classification and 5.9% on AUROC for OOD detection on TinyImageNet dataset. The source codes are publicly available at https://github.com/rain305f/OSP." @default.
- W4386083105 created "2023-08-23" @default.
- W4386083105 creator A5008752777 @default.
- W4386083105 creator A5010479652 @default.
- W4386083105 creator A5022256556 @default.
- W4386083105 creator A5054226277 @default.
- W4386083105 creator A5079594267 @default.
- W4386083105 creator A5087192063 @default.
- W4386083105 date "2023-06-01" @default.
- W4386083105 modified "2023-09-26" @default.
- W4386083105 title "Out-of-Distributed Semantic Pruning for Robust Semi-Supervised Learning" @default.
- W4386083105 cites W2108598243 @default.
- W4386083105 cites W2133059825 @default.
- W4386083105 cites W2618530766 @default.
- W4386083105 cites W2943766732 @default.
- W4386083105 cites W2949736877 @default.
- W4386083105 cites W2962858109 @default.
- W4386083105 cites W2964159205 @default.
- W4386083105 cites W2979805229 @default.
- W4386083105 cites W2997131443 @default.
- W4386083105 cites W3035235554 @default.
- W4386083105 cites W3094146654 @default.
- W4386083105 cites W3094447715 @default.
- W4386083105 cites W3174697615 @default.
- W4386083105 cites W3191916981 @default.
- W4386083105 cites W4225343834 @default.
- W4386083105 cites W4293057377 @default.
- W4386083105 cites W4313138032 @default.
- W4386083105 doi "https://doi.org/10.1109/cvpr52729.2023.02284" @default.
- W4386083105 hasPublicationYear "2023" @default.
- W4386083105 type Work @default.
- W4386083105 citedByCount "0" @default.
- W4386083105 crossrefType "proceedings-article" @default.
- W4386083105 hasAuthorship W4386083105A5008752777 @default.
- W4386083105 hasAuthorship W4386083105A5010479652 @default.
- W4386083105 hasAuthorship W4386083105A5022256556 @default.
- W4386083105 hasAuthorship W4386083105A5054226277 @default.
- W4386083105 hasAuthorship W4386083105A5079594267 @default.
- W4386083105 hasAuthorship W4386083105A5087192063 @default.
- W4386083105 hasConcept C105795698 @default.
- W4386083105 hasConcept C108010975 @default.
- W4386083105 hasConcept C119857082 @default.
- W4386083105 hasConcept C153180895 @default.
- W4386083105 hasConcept C154945302 @default.
- W4386083105 hasConcept C165064840 @default.
- W4386083105 hasConcept C17137986 @default.
- W4386083105 hasConcept C184337299 @default.
- W4386083105 hasConcept C199360897 @default.
- W4386083105 hasConcept C2524010 @default.
- W4386083105 hasConcept C2778493491 @default.
- W4386083105 hasConcept C2781122975 @default.
- W4386083105 hasConcept C33923547 @default.
- W4386083105 hasConcept C41008148 @default.
- W4386083105 hasConcept C6557445 @default.
- W4386083105 hasConcept C86803240 @default.
- W4386083105 hasConceptScore W4386083105C105795698 @default.
- W4386083105 hasConceptScore W4386083105C108010975 @default.
- W4386083105 hasConceptScore W4386083105C119857082 @default.
- W4386083105 hasConceptScore W4386083105C153180895 @default.
- W4386083105 hasConceptScore W4386083105C154945302 @default.
- W4386083105 hasConceptScore W4386083105C165064840 @default.
- W4386083105 hasConceptScore W4386083105C17137986 @default.
- W4386083105 hasConceptScore W4386083105C184337299 @default.
- W4386083105 hasConceptScore W4386083105C199360897 @default.
- W4386083105 hasConceptScore W4386083105C2524010 @default.
- W4386083105 hasConceptScore W4386083105C2778493491 @default.
- W4386083105 hasConceptScore W4386083105C2781122975 @default.
- W4386083105 hasConceptScore W4386083105C33923547 @default.
- W4386083105 hasConceptScore W4386083105C41008148 @default.
- W4386083105 hasConceptScore W4386083105C6557445 @default.
- W4386083105 hasConceptScore W4386083105C86803240 @default.
- W4386083105 hasFunder F4320321001 @default.
- W4386083105 hasFunder F4320321921 @default.
- W4386083105 hasFunder F4320335777 @default.
- W4386083105 hasLocation W43860831051 @default.
- W4386083105 hasOpenAccess W4386083105 @default.
- W4386083105 hasPrimaryLocation W43860831051 @default.
- W4386083105 hasRelatedWork W1973091277 @default.
- W4386083105 hasRelatedWork W2117181705 @default.
- W4386083105 hasRelatedWork W2380389143 @default.
- W4386083105 hasRelatedWork W2943995822 @default.
- W4386083105 hasRelatedWork W2982420602 @default.
- W4386083105 hasRelatedWork W3091418883 @default.
- W4386083105 hasRelatedWork W3188199900 @default.
- W4386083105 hasRelatedWork W3201762449 @default.
- W4386083105 hasRelatedWork W4323076479 @default.
- W4386083105 hasRelatedWork W4375869227 @default.
- W4386083105 isParatext "false" @default.
- W4386083105 isRetracted "false" @default.
- W4386083105 workType "article" @default.