Matches in SemOpenAlex for { <https://semopenalex.org/work/W3204341018> ?p ?o ?g. }
Showing items 1 to 73 of
73
with 100 items per page.
- W3204341018 abstract "A myriad of recent breakthroughs in hand-crafted neural architectures for visual recognition have highlighted the urgent need to explore hybrid architectures consisting of diversified building blocks. Meanwhile, neural architecture search methods are surging with an expectation to reduce human efforts. However, whether NAS methods can efficiently and effectively handle diversified search spaces with disparate candidates (e.g. CNNs and transformers) is still an open question. In this work, we present Block-wisely Self-supervised Neural Architecture Search (BossNAS), an unsupervised NAS method that addresses the problem of inaccurate architecture rating caused by large weight-sharing space and biased supervision in previous methods. More specifically, we factorize the search space into blocks and utilize a novel self-supervised training scheme, named ensemble bootstrapping, to train each block separately before searching them as a whole towards the population center. Additionally, we present HyTra search space, a fabric-like hybrid CNN-transformer search space with searchable down-sampling positions. On this challenging search space, our searched model, BossNet-T, achieves up to 82.5% accuracy on ImageNet, surpassing EfficientNet by 2.4% with comparable compute time. Moreover, our method achieves superior architecture rating accuracy with 0.78 and 0.76 Spearman correlation on the canonical MBConv search space with ImageNet and on NATS-Bench size search space with CIFAR-100, respectively, surpassing state-of-the-art NAS methods. Code: https://github.com/changlin31/BossNAS" @default.
- W3204341018 created "2021-10-11" @default.
- W3204341018 creator A5024168418 @default.
- W3204341018 creator A5034967388 @default.
- W3204341018 creator A5040040696 @default.
- W3204341018 creator A5047878798 @default.
- W3204341018 creator A5048957271 @default.
- W3204341018 creator A5052611320 @default.
- W3204341018 creator A5081233515 @default.
- W3204341018 date "2021-03-23" @default.
- W3204341018 modified "2023-10-12" @default.
- W3204341018 title "BossNAS: Exploring Hybrid CNN-transformers with Block-wisely Self-supervised Neural Architecture Search" @default.
- W3204341018 doi "https://doi.org/10.48550/arxiv.2103.12424" @default.
- W3204341018 hasPublicationYear "2021" @default.
- W3204341018 type Work @default.
- W3204341018 sameAs 3204341018 @default.
- W3204341018 citedByCount "1" @default.
- W3204341018 crossrefType "posted-content" @default.
- W3204341018 hasAuthorship W3204341018A5024168418 @default.
- W3204341018 hasAuthorship W3204341018A5034967388 @default.
- W3204341018 hasAuthorship W3204341018A5040040696 @default.
- W3204341018 hasAuthorship W3204341018A5047878798 @default.
- W3204341018 hasAuthorship W3204341018A5048957271 @default.
- W3204341018 hasAuthorship W3204341018A5052611320 @default.
- W3204341018 hasAuthorship W3204341018A5081233515 @default.
- W3204341018 hasBestOaLocation W32043410181 @default.
- W3204341018 hasConcept C119599485 @default.
- W3204341018 hasConcept C119857082 @default.
- W3204341018 hasConcept C123657996 @default.
- W3204341018 hasConcept C127413603 @default.
- W3204341018 hasConcept C142362112 @default.
- W3204341018 hasConcept C153349607 @default.
- W3204341018 hasConcept C154945302 @default.
- W3204341018 hasConcept C165801399 @default.
- W3204341018 hasConcept C2524010 @default.
- W3204341018 hasConcept C2777210771 @default.
- W3204341018 hasConcept C33923547 @default.
- W3204341018 hasConcept C41008148 @default.
- W3204341018 hasConcept C50644808 @default.
- W3204341018 hasConcept C66322947 @default.
- W3204341018 hasConcept C81363708 @default.
- W3204341018 hasConceptScore W3204341018C119599485 @default.
- W3204341018 hasConceptScore W3204341018C119857082 @default.
- W3204341018 hasConceptScore W3204341018C123657996 @default.
- W3204341018 hasConceptScore W3204341018C127413603 @default.
- W3204341018 hasConceptScore W3204341018C142362112 @default.
- W3204341018 hasConceptScore W3204341018C153349607 @default.
- W3204341018 hasConceptScore W3204341018C154945302 @default.
- W3204341018 hasConceptScore W3204341018C165801399 @default.
- W3204341018 hasConceptScore W3204341018C2524010 @default.
- W3204341018 hasConceptScore W3204341018C2777210771 @default.
- W3204341018 hasConceptScore W3204341018C33923547 @default.
- W3204341018 hasConceptScore W3204341018C41008148 @default.
- W3204341018 hasConceptScore W3204341018C50644808 @default.
- W3204341018 hasConceptScore W3204341018C66322947 @default.
- W3204341018 hasConceptScore W3204341018C81363708 @default.
- W3204341018 hasLocation W32043410181 @default.
- W3204341018 hasOpenAccess W3204341018 @default.
- W3204341018 hasPrimaryLocation W32043410181 @default.
- W3204341018 hasRelatedWork W2337926734 @default.
- W3204341018 hasRelatedWork W2799614062 @default.
- W3204341018 hasRelatedWork W2978290780 @default.
- W3204341018 hasRelatedWork W3027997911 @default.
- W3204341018 hasRelatedWork W3052481912 @default.
- W3204341018 hasRelatedWork W3173182854 @default.
- W3204341018 hasRelatedWork W4281780675 @default.
- W3204341018 hasRelatedWork W4285586943 @default.
- W3204341018 hasRelatedWork W4287776258 @default.
- W3204341018 hasRelatedWork W3009789068 @default.
- W3204341018 isParatext "false" @default.
- W3204341018 isRetracted "false" @default.
- W3204341018 magId "3204341018" @default.
- W3204341018 workType "article" @default.