Matches in SemOpenAlex for { <https://semopenalex.org/work/W4312692019> ?p ?o ?g. }
Showing items 1 to 88 of
88
with 100 items per page.
- W4312692019 abstract "The Sinkhorn operator has recently experienced a surge of popularity in computer vision and related fields. One major reason is its ease of integration into deep learning frameworks. To allow for an efficient training of respective neural networks, we propose an algorithm that obtains analytical gradients of a Sinkhorn layer via implicit differentiation. In comparison to prior work, our framework is based on the most general formulation of the Sinkhorn operator. It allows for any type of loss function, while both the target capacities and cost matrices are differentiated jointly. We further construct error bounds of the resulting algorithm for approximate inputs. Finally, we demonstrate that for a number of applications, simply replacing automatic differentiation with our algorithm directly improves the stability and accuracy of the obtained gradients. Moreover, we show that it is computationally more efficient, particularly when resources like GPU memory are scarce. <sup xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>1</sup> <sup xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>1</sup> Our implementation is available under the following link: https://github.com/marvin-eisenberger/implicit-sinkhorn" @default.
- W4312692019 created "2023-01-05" @default.
- W4312692019 creator A5032671136 @default.
- W4312692019 creator A5045111379 @default.
- W4312692019 creator A5051910500 @default.
- W4312692019 creator A5087710605 @default.
- W4312692019 creator A5090283032 @default.
- W4312692019 date "2022-06-01" @default.
- W4312692019 modified "2023-09-30" @default.
- W4312692019 title "A Unified Framework for Implicit Sinkhorn Differentiation" @default.
- W4312692019 cites W1585160083 @default.
- W4312692019 cites W2009172320 @default.
- W4312692019 cites W2222512263 @default.
- W4312692019 cites W2509351257 @default.
- W4312692019 cites W2539033431 @default.
- W4312692019 cites W2606014079 @default.
- W4312692019 cites W3007724151 @default.
- W4312692019 cites W3013243617 @default.
- W4312692019 cites W3033420020 @default.
- W4312692019 cites W3034275286 @default.
- W4312692019 cites W3102246635 @default.
- W4312692019 cites W3109766878 @default.
- W4312692019 cites W3159359216 @default.
- W4312692019 cites W4206471589 @default.
- W4312692019 cites W4210706564 @default.
- W4312692019 doi "https://doi.org/10.1109/cvpr52688.2022.00059" @default.
- W4312692019 hasPublicationYear "2022" @default.
- W4312692019 type Work @default.
- W4312692019 citedByCount "2" @default.
- W4312692019 countsByYear W43126920192023 @default.
- W4312692019 crossrefType "proceedings-article" @default.
- W4312692019 hasAuthorship W4312692019A5032671136 @default.
- W4312692019 hasAuthorship W4312692019A5045111379 @default.
- W4312692019 hasAuthorship W4312692019A5051910500 @default.
- W4312692019 hasAuthorship W4312692019A5087710605 @default.
- W4312692019 hasAuthorship W4312692019A5090283032 @default.
- W4312692019 hasBestOaLocation W43126920192 @default.
- W4312692019 hasConcept C104317684 @default.
- W4312692019 hasConcept C112972136 @default.
- W4312692019 hasConcept C11413529 @default.
- W4312692019 hasConcept C119857082 @default.
- W4312692019 hasConcept C133512626 @default.
- W4312692019 hasConcept C154945302 @default.
- W4312692019 hasConcept C158448853 @default.
- W4312692019 hasConcept C17020691 @default.
- W4312692019 hasConcept C178790620 @default.
- W4312692019 hasConcept C185592680 @default.
- W4312692019 hasConcept C2779227376 @default.
- W4312692019 hasConcept C41008148 @default.
- W4312692019 hasConcept C45374587 @default.
- W4312692019 hasConcept C50644808 @default.
- W4312692019 hasConcept C55493867 @default.
- W4312692019 hasConcept C80444323 @default.
- W4312692019 hasConcept C86339819 @default.
- W4312692019 hasConceptScore W4312692019C104317684 @default.
- W4312692019 hasConceptScore W4312692019C112972136 @default.
- W4312692019 hasConceptScore W4312692019C11413529 @default.
- W4312692019 hasConceptScore W4312692019C119857082 @default.
- W4312692019 hasConceptScore W4312692019C133512626 @default.
- W4312692019 hasConceptScore W4312692019C154945302 @default.
- W4312692019 hasConceptScore W4312692019C158448853 @default.
- W4312692019 hasConceptScore W4312692019C17020691 @default.
- W4312692019 hasConceptScore W4312692019C178790620 @default.
- W4312692019 hasConceptScore W4312692019C185592680 @default.
- W4312692019 hasConceptScore W4312692019C2779227376 @default.
- W4312692019 hasConceptScore W4312692019C41008148 @default.
- W4312692019 hasConceptScore W4312692019C45374587 @default.
- W4312692019 hasConceptScore W4312692019C50644808 @default.
- W4312692019 hasConceptScore W4312692019C55493867 @default.
- W4312692019 hasConceptScore W4312692019C80444323 @default.
- W4312692019 hasConceptScore W4312692019C86339819 @default.
- W4312692019 hasLocation W43126920191 @default.
- W4312692019 hasLocation W43126920192 @default.
- W4312692019 hasOpenAccess W4312692019 @default.
- W4312692019 hasPrimaryLocation W43126920191 @default.
- W4312692019 hasRelatedWork W1575012072 @default.
- W4312692019 hasRelatedWork W1989482460 @default.
- W4312692019 hasRelatedWork W2081174236 @default.
- W4312692019 hasRelatedWork W2093594655 @default.
- W4312692019 hasRelatedWork W2375380208 @default.
- W4312692019 hasRelatedWork W2380313759 @default.
- W4312692019 hasRelatedWork W2386387936 @default.
- W4312692019 hasRelatedWork W2386767533 @default.
- W4312692019 hasRelatedWork W2392110728 @default.
- W4312692019 hasRelatedWork W1629725936 @default.
- W4312692019 isParatext "false" @default.
- W4312692019 isRetracted "false" @default.
- W4312692019 workType "article" @default.