Matches in SemOpenAlex for { <https://semopenalex.org/work/W3163602117> ?p ?o ?g. }
- W3163602117 abstract "Recent self-supervised methods for image representation learning are based on maximizing the agreement between embedding vectors from different views of the same image. A trivial solution is obtained when the encoder outputs constant vectors. This collapse problem is often avoided through implicit biases in the learning architecture, that often lack a clear justification or interpretation. In this paper, we introduce VICReg (Variance-Invariance-Covariance Regularization), a method that explicitly avoids the collapse problem with a simple regularization term on the variance of the embeddings along each dimension individually. VICReg combines the variance term with a decorrelation mechanism based on redundancy reduction and covariance regularization, and achieves results on par with the state of the art on several downstream tasks. In addition, we show that incorporating our new variance term into other methods helps stabilize the training and leads to performance improvements." @default.
- W3163602117 created "2021-05-24" @default.
- W3163602117 creator A5001226970 @default.
- W3163602117 creator A5033229022 @default.
- W3163602117 creator A5039530212 @default.
- W3163602117 date "2022-04-25" @default.
- W3163602117 modified "2023-10-17" @default.
- W3163602117 title "VICReg: Variance-Invariance-Covariance Regularization for Self-Supervised Learning" @default.
- W3163602117 cites W1821462560 @default.
- W3163602117 cites W1836465849 @default.
- W3163602117 cites W1861492603 @default.
- W3163602117 cites W2031489346 @default.
- W3163602117 cites W2052666245 @default.
- W3163602117 cites W2108598243 @default.
- W3163602117 cites W2118585731 @default.
- W3163602117 cites W2127589108 @default.
- W3163602117 cites W2134670479 @default.
- W3163602117 cites W2138621090 @default.
- W3163602117 cites W2157364932 @default.
- W3163602117 cites W2158131535 @default.
- W3163602117 cites W2194775991 @default.
- W3163602117 cites W2401231614 @default.
- W3163602117 cites W2516580127 @default.
- W3163602117 cites W2549139847 @default.
- W3163602117 cites W2613718673 @default.
- W3163602117 cites W2622263826 @default.
- W3163602117 cites W2757910899 @default.
- W3163602117 cites W2797977484 @default.
- W3163602117 cites W2798991696 @default.
- W3163602117 cites W2807205843 @default.
- W3163602117 cites W2842511635 @default.
- W3163602117 cites W2887997457 @default.
- W3163602117 cites W2941964676 @default.
- W3163602117 cites W2944828972 @default.
- W3163602117 cites W2949517790 @default.
- W3163602117 cites W2952122856 @default.
- W3163602117 cites W2962852342 @default.
- W3163602117 cites W2962877362 @default.
- W3163602117 cites W2963263347 @default.
- W3163602117 cites W2964074409 @default.
- W3163602117 cites W2971155163 @default.
- W3163602117 cites W2979579363 @default.
- W3163602117 cites W2987741655 @default.
- W3163602117 cites W2995181141 @default.
- W3163602117 cites W2998388430 @default.
- W3163602117 cites W3005680577 @default.
- W3163602117 cites W3009561768 @default.
- W3163602117 cites W3018265077 @default.
- W3163602117 cites W3022061250 @default.
- W3163602117 cites W3034345981 @default.
- W3163602117 cites W3034781633 @default.
- W3163602117 cites W3035058308 @default.
- W3163602117 cites W3035524453 @default.
- W3163602117 cites W3041919418 @default.
- W3163602117 cites W3093929102 @default.
- W3163602117 cites W3095121901 @default.
- W3163602117 cites W3100859887 @default.
- W3163602117 cites W3101821705 @default.
- W3163602117 cites W3106428938 @default.
- W3163602117 cites W3116557712 @default.
- W3163602117 cites W3119786062 @default.
- W3163602117 cites W3122325173 @default.
- W3163602117 cites W3131573008 @default.
- W3163602117 cites W3134652006 @default.
- W3163602117 cites W3171007011 @default.
- W3163602117 cites W343636949 @default.
- W3163602117 hasPublicationYear "2022" @default.
- W3163602117 type Work @default.
- W3163602117 sameAs 3163602117 @default.
- W3163602117 citedByCount "19" @default.
- W3163602117 countsByYear W31636021172021 @default.
- W3163602117 crossrefType "proceedings-article" @default.
- W3163602117 hasAuthorship W3163602117A5001226970 @default.
- W3163602117 hasAuthorship W3163602117A5033229022 @default.
- W3163602117 hasAuthorship W3163602117A5039530212 @default.
- W3163602117 hasBestOaLocation W31636021171 @default.
- W3163602117 hasConcept C105795698 @default.
- W3163602117 hasConcept C11413529 @default.
- W3163602117 hasConcept C119340705 @default.
- W3163602117 hasConcept C119857082 @default.
- W3163602117 hasConcept C121955636 @default.
- W3163602117 hasConcept C144133560 @default.
- W3163602117 hasConcept C153180895 @default.
- W3163602117 hasConcept C154945302 @default.
- W3163602117 hasConcept C178650346 @default.
- W3163602117 hasConcept C196083921 @default.
- W3163602117 hasConcept C2776135515 @default.
- W3163602117 hasConcept C28826006 @default.
- W3163602117 hasConcept C33923547 @default.
- W3163602117 hasConcept C41008148 @default.
- W3163602117 hasConceptScore W3163602117C105795698 @default.
- W3163602117 hasConceptScore W3163602117C11413529 @default.
- W3163602117 hasConceptScore W3163602117C119340705 @default.
- W3163602117 hasConceptScore W3163602117C119857082 @default.
- W3163602117 hasConceptScore W3163602117C121955636 @default.
- W3163602117 hasConceptScore W3163602117C144133560 @default.
- W3163602117 hasConceptScore W3163602117C153180895 @default.
- W3163602117 hasConceptScore W3163602117C154945302 @default.
- W3163602117 hasConceptScore W3163602117C178650346 @default.
- W3163602117 hasConceptScore W3163602117C196083921 @default.