Matches in SemOpenAlex for { <https://semopenalex.org/work/W3158424761> ?p ?o ?g. }
- W3158424761 abstract "In self-supervised representation learning, a common idea behind most of the state-of-the-art approaches is to enforce the robustness of the representations to predefined augmentations. A potential issue of this idea is the existence of completely collapsed solutions (i.e., constant features), which are typically avoided implicitly by carefully chosen implementation details. In this work, we study a relatively concise framework containing the most common components from recent approaches. We verify the existence of complete collapse and discover another reachable collapse pattern that is usually overlooked, namely dimensional collapse. We connect dimensional collapse with strong correlations between axes and consider such connection as a strong motivation for feature decorrelation (i.e., standardizing the covariance matrix). The gains from feature decorrelation are verified empirically to highlight the importance and the potential of this insight." @default.
- W3158424761 created "2021-05-10" @default.
- W3158424761 creator A5033632697 @default.
- W3158424761 creator A5041159177 @default.
- W3158424761 creator A5044430605 @default.
- W3158424761 creator A5048891976 @default.
- W3158424761 creator A5059038346 @default.
- W3158424761 creator A5062587172 @default.
- W3158424761 date "2021-05-02" @default.
- W3158424761 modified "2023-09-26" @default.
- W3158424761 title "On Feature Decorrelation in Self-Supervised Learning" @default.
- W3158424761 cites W1836465849 @default.
- W3158424761 cites W2117539524 @default.
- W3158424761 cites W2137234026 @default.
- W3158424761 cites W2194775991 @default.
- W3158424761 cites W2518108298 @default.
- W3158424761 cites W2533523411 @default.
- W3158424761 cites W2583938035 @default.
- W3158424761 cites W2622263826 @default.
- W3158424761 cites W2757910899 @default.
- W3158424761 cites W2798991696 @default.
- W3158424761 cites W2842511635 @default.
- W3158424761 cites W2883725317 @default.
- W3158424761 cites W2887997457 @default.
- W3158424761 cites W2896457183 @default.
- W3158424761 cites W2917375850 @default.
- W3158424761 cites W2962684187 @default.
- W3158424761 cites W2963743626 @default.
- W3158424761 cites W2965954137 @default.
- W3158424761 cites W2979475574 @default.
- W3158424761 cites W2987741655 @default.
- W3158424761 cites W2995181141 @default.
- W3158424761 cites W2998388430 @default.
- W3158424761 cites W3009561768 @default.
- W3158424761 cites W3022061250 @default.
- W3158424761 cites W3034978746 @default.
- W3158424761 cites W3035058308 @default.
- W3158424761 cites W3035060554 @default.
- W3158424761 cites W3035524453 @default.
- W3158424761 cites W3041919418 @default.
- W3158424761 cites W3061525482 @default.
- W3158424761 cites W3093929102 @default.
- W3158424761 cites W3095121901 @default.
- W3158424761 cites W3099331343 @default.
- W3158424761 cites W3106005682 @default.
- W3158424761 cites W3106310609 @default.
- W3158424761 cites W3106428938 @default.
- W3158424761 cites W3106485021 @default.
- W3158424761 cites W3106743555 @default.
- W3158424761 cites W3107668149 @default.
- W3158424761 cites W3113997557 @default.
- W3158424761 cites W3118062200 @default.
- W3158424761 cites W3118608800 @default.
- W3158424761 cites W3123939835 @default.
- W3158424761 cites W3131573008 @default.
- W3158424761 cites W3133014250 @default.
- W3158424761 cites W3149106692 @default.
- W3158424761 cites W3166500718 @default.
- W3158424761 cites W2551176409 @default.
- W3158424761 doi "https://doi.org/10.48550/arxiv.2105.00470" @default.
- W3158424761 hasPublicationYear "2021" @default.
- W3158424761 type Work @default.
- W3158424761 sameAs 3158424761 @default.
- W3158424761 citedByCount "4" @default.
- W3158424761 countsByYear W31584247612021 @default.
- W3158424761 crossrefType "posted-content" @default.
- W3158424761 hasAuthorship W3158424761A5033632697 @default.
- W3158424761 hasAuthorship W3158424761A5041159177 @default.
- W3158424761 hasAuthorship W3158424761A5044430605 @default.
- W3158424761 hasAuthorship W3158424761A5048891976 @default.
- W3158424761 hasAuthorship W3158424761A5059038346 @default.
- W3158424761 hasAuthorship W3158424761A5062587172 @default.
- W3158424761 hasBestOaLocation W31584247611 @default.
- W3158424761 hasConcept C104317684 @default.
- W3158424761 hasConcept C105795698 @default.
- W3158424761 hasConcept C11413529 @default.
- W3158424761 hasConcept C119857082 @default.
- W3158424761 hasConcept C138885662 @default.
- W3158424761 hasConcept C153180895 @default.
- W3158424761 hasConcept C154945302 @default.
- W3158424761 hasConcept C17744445 @default.
- W3158424761 hasConcept C177860922 @default.
- W3158424761 hasConcept C178650346 @default.
- W3158424761 hasConcept C185592680 @default.
- W3158424761 hasConcept C199539241 @default.
- W3158424761 hasConcept C2776359362 @default.
- W3158424761 hasConcept C2776401178 @default.
- W3158424761 hasConcept C33923547 @default.
- W3158424761 hasConcept C41008148 @default.
- W3158424761 hasConcept C41895202 @default.
- W3158424761 hasConcept C55493867 @default.
- W3158424761 hasConcept C59404180 @default.
- W3158424761 hasConcept C63479239 @default.
- W3158424761 hasConcept C80444323 @default.
- W3158424761 hasConcept C94625758 @default.
- W3158424761 hasConceptScore W3158424761C104317684 @default.
- W3158424761 hasConceptScore W3158424761C105795698 @default.
- W3158424761 hasConceptScore W3158424761C11413529 @default.
- W3158424761 hasConceptScore W3158424761C119857082 @default.
- W3158424761 hasConceptScore W3158424761C138885662 @default.