Matches in SemOpenAlex for { <https://semopenalex.org/work/W3208452612> ?p ?o ?g. }
- W3208452612 abstract "The discovery of sparse subnetworks that are able to perform as well as full models has found broad applied and theoretical interest. While many pruning methods have been developed to this end, the naive approach of removing parameters based on their magnitude has been found to be as robust as more complex, state-of-the-art algorithms. The lack of theory behind magnitude pruning's success, especially pre-convergence, and its relation to other pruning methods, such as gradient based pruning, are outstanding open questions in the field that are in need of being addressed. We make use of recent advances in dynamical systems theory, namely Koopman operator theory, to define a new class of theoretically motivated pruning algorithms. We show that these algorithms can be equivalent to magnitude and gradient based pruning, unifying these seemingly disparate methods, and that they can be used to shed light on magnitude pruning's performance during early training." @default.
- W3208452612 created "2021-11-08" @default.
- W3208452612 creator A5014334832 @default.
- W3208452612 creator A5020979758 @default.
- W3208452612 creator A5036566464 @default.
- W3208452612 creator A5064887522 @default.
- W3208452612 creator A5067108617 @default.
- W3208452612 date "2021-10-28" @default.
- W3208452612 modified "2023-09-26" @default.
- W3208452612 title "An Operator Theoretic Perspective on Pruning Deep Neural Networks" @default.
- W3208452612 cites W1997054581 @default.
- W3208452612 cites W2013825190 @default.
- W3208452612 cites W2014356541 @default.
- W3208452612 cites W2025713906 @default.
- W3208452612 cites W2060806808 @default.
- W3208452612 cites W2067035505 @default.
- W3208452612 cites W2094686470 @default.
- W3208452612 cites W2097533491 @default.
- W3208452612 cites W2114766824 @default.
- W3208452612 cites W2114876173 @default.
- W3208452612 cites W2125389748 @default.
- W3208452612 cites W2164954534 @default.
- W3208452612 cites W2904243021 @default.
- W3208452612 cites W2911040189 @default.
- W3208452612 cites W2963247446 @default.
- W3208452612 cites W2963674932 @default.
- W3208452612 cites W2963813662 @default.
- W3208452612 cites W2971127900 @default.
- W3208452612 cites W2973022759 @default.
- W3208452612 cites W2991844785 @default.
- W3208452612 cites W2996577930 @default.
- W3208452612 cites W3010156474 @default.
- W3208452612 cites W3035081900 @default.
- W3208452612 cites W3036818215 @default.
- W3208452612 cites W3037500280 @default.
- W3208452612 cites W3042900599 @default.
- W3208452612 cites W3084942782 @default.
- W3208452612 cites W3099803807 @default.
- W3208452612 cites W3100771563 @default.
- W3208452612 cites W3120244898 @default.
- W3208452612 cites W3132732981 @default.
- W3208452612 hasPublicationYear "2021" @default.
- W3208452612 type Work @default.
- W3208452612 sameAs 3208452612 @default.
- W3208452612 citedByCount "0" @default.
- W3208452612 crossrefType "posted-content" @default.
- W3208452612 hasAuthorship W3208452612A5014334832 @default.
- W3208452612 hasAuthorship W3208452612A5020979758 @default.
- W3208452612 hasAuthorship W3208452612A5036566464 @default.
- W3208452612 hasAuthorship W3208452612A5064887522 @default.
- W3208452612 hasAuthorship W3208452612A5067108617 @default.
- W3208452612 hasConcept C104317684 @default.
- W3208452612 hasConcept C108010975 @default.
- W3208452612 hasConcept C11413529 @default.
- W3208452612 hasConcept C119857082 @default.
- W3208452612 hasConcept C12713177 @default.
- W3208452612 hasConcept C154945302 @default.
- W3208452612 hasConcept C158448853 @default.
- W3208452612 hasConcept C162324750 @default.
- W3208452612 hasConcept C17020691 @default.
- W3208452612 hasConcept C185592680 @default.
- W3208452612 hasConcept C2777212361 @default.
- W3208452612 hasConcept C2777303404 @default.
- W3208452612 hasConcept C41008148 @default.
- W3208452612 hasConcept C50522688 @default.
- W3208452612 hasConcept C50644808 @default.
- W3208452612 hasConcept C55493867 @default.
- W3208452612 hasConcept C6557445 @default.
- W3208452612 hasConcept C86339819 @default.
- W3208452612 hasConcept C86803240 @default.
- W3208452612 hasConceptScore W3208452612C104317684 @default.
- W3208452612 hasConceptScore W3208452612C108010975 @default.
- W3208452612 hasConceptScore W3208452612C11413529 @default.
- W3208452612 hasConceptScore W3208452612C119857082 @default.
- W3208452612 hasConceptScore W3208452612C12713177 @default.
- W3208452612 hasConceptScore W3208452612C154945302 @default.
- W3208452612 hasConceptScore W3208452612C158448853 @default.
- W3208452612 hasConceptScore W3208452612C162324750 @default.
- W3208452612 hasConceptScore W3208452612C17020691 @default.
- W3208452612 hasConceptScore W3208452612C185592680 @default.
- W3208452612 hasConceptScore W3208452612C2777212361 @default.
- W3208452612 hasConceptScore W3208452612C2777303404 @default.
- W3208452612 hasConceptScore W3208452612C41008148 @default.
- W3208452612 hasConceptScore W3208452612C50522688 @default.
- W3208452612 hasConceptScore W3208452612C50644808 @default.
- W3208452612 hasConceptScore W3208452612C55493867 @default.
- W3208452612 hasConceptScore W3208452612C6557445 @default.
- W3208452612 hasConceptScore W3208452612C86339819 @default.
- W3208452612 hasConceptScore W3208452612C86803240 @default.
- W3208452612 hasLocation W32084526121 @default.
- W3208452612 hasOpenAccess W3208452612 @default.
- W3208452612 hasPrimaryLocation W32084526121 @default.
- W3208452612 hasRelatedWork W1595732048 @default.
- W3208452612 hasRelatedWork W1890558254 @default.
- W3208452612 hasRelatedWork W2081851784 @default.
- W3208452612 hasRelatedWork W2105459687 @default.
- W3208452612 hasRelatedWork W2106411961 @default.
- W3208452612 hasRelatedWork W2136136954 @default.
- W3208452612 hasRelatedWork W2198787668 @default.
- W3208452612 hasRelatedWork W2248157934 @default.