Matches in SemOpenAlex for { <https://semopenalex.org/work/W3209835134> ?p ?o ?g. }
- W3209835134 abstract "An often unquestioned assumption underlying most current federated learning algorithms is that all the participants use identical model architectures. In this work, we initiate a theoretical study of model agnostic communication protocols which would allow data holders (agents) using different models to collaborate with each other and perform federated learning. We focus on the setting where the two agents are attempting to perform kernel regression using different kernels (and hence have different models). Our study yields a surprising result -- the most natural algorithm of using alternating knowledge distillation (AKD) imposes overly strong regularization and may lead to severe under-fitting. Our theory also shows an interesting connection between AKD and the alternating projection algorithm for finding intersection of sets. Leveraging this connection, we propose a new algorithms which improve upon AKD. Our theoretical predictions also closely match real world experiments using neural networks. Thus, our work proposes a rich yet tractable framework for analyzing and developing new practical model agnostic federated learning algorithms." @default.
- W3209835134 created "2021-11-08" @default.
- W3209835134 creator A5026569995 @default.
- W3209835134 creator A5054671348 @default.
- W3209835134 date "2021-10-28" @default.
- W3209835134 modified "2023-10-01" @default.
- W3209835134 title "Towards Model Agnostic Federated Learning Using Knowledge Distillation." @default.
- W3209835134 cites W1540155273 @default.
- W3209835134 cites W1542938076 @default.
- W3209835134 cites W1821462560 @default.
- W3209835134 cites W1986280275 @default.
- W3209835134 cites W2158902838 @default.
- W3209835134 cites W2541884796 @default.
- W3209835134 cites W2619516334 @default.
- W3209835134 cites W2620998106 @default.
- W3209835134 cites W2739879705 @default.
- W3209835134 cites W2787017828 @default.
- W3209835134 cites W2963318081 @default.
- W3209835134 cites W2981206218 @default.
- W3209835134 cites W2981738522 @default.
- W3209835134 cites W2995022099 @default.
- W3209835134 cites W3004809259 @default.
- W3209835134 cites W3007548213 @default.
- W3209835134 cites W3008187686 @default.
- W3209835134 cites W3012968339 @default.
- W3209835134 cites W3022064003 @default.
- W3209835134 cites W3028069987 @default.
- W3209835134 cites W3035453001 @default.
- W3209835134 cites W3047989515 @default.
- W3209835134 cites W3091829090 @default.
- W3209835134 cites W3096572172 @default.
- W3209835134 cites W3098105997 @default.
- W3209835134 cites W3103361051 @default.
- W3209835134 cites W3108358516 @default.
- W3209835134 cites W3113303810 @default.
- W3209835134 cites W3160036634 @default.
- W3209835134 cites W3165285531 @default.
- W3209835134 cites W3168256142 @default.
- W3209835134 cites W3178336997 @default.
- W3209835134 cites W3200971224 @default.
- W3209835134 cites W3211115474 @default.
- W3209835134 hasPublicationYear "2021" @default.
- W3209835134 type Work @default.
- W3209835134 sameAs 3209835134 @default.
- W3209835134 citedByCount "0" @default.
- W3209835134 crossrefType "posted-content" @default.
- W3209835134 hasAuthorship W3209835134A5026569995 @default.
- W3209835134 hasAuthorship W3209835134A5054671348 @default.
- W3209835134 hasConcept C114614502 @default.
- W3209835134 hasConcept C119857082 @default.
- W3209835134 hasConcept C120665830 @default.
- W3209835134 hasConcept C121332964 @default.
- W3209835134 hasConcept C127413603 @default.
- W3209835134 hasConcept C13355873 @default.
- W3209835134 hasConcept C146978453 @default.
- W3209835134 hasConcept C154945302 @default.
- W3209835134 hasConcept C178790620 @default.
- W3209835134 hasConcept C185592680 @default.
- W3209835134 hasConcept C192209626 @default.
- W3209835134 hasConcept C204030448 @default.
- W3209835134 hasConcept C2524010 @default.
- W3209835134 hasConcept C2776135515 @default.
- W3209835134 hasConcept C33923547 @default.
- W3209835134 hasConcept C41008148 @default.
- W3209835134 hasConcept C64543145 @default.
- W3209835134 hasConcept C74193536 @default.
- W3209835134 hasConceptScore W3209835134C114614502 @default.
- W3209835134 hasConceptScore W3209835134C119857082 @default.
- W3209835134 hasConceptScore W3209835134C120665830 @default.
- W3209835134 hasConceptScore W3209835134C121332964 @default.
- W3209835134 hasConceptScore W3209835134C127413603 @default.
- W3209835134 hasConceptScore W3209835134C13355873 @default.
- W3209835134 hasConceptScore W3209835134C146978453 @default.
- W3209835134 hasConceptScore W3209835134C154945302 @default.
- W3209835134 hasConceptScore W3209835134C178790620 @default.
- W3209835134 hasConceptScore W3209835134C185592680 @default.
- W3209835134 hasConceptScore W3209835134C192209626 @default.
- W3209835134 hasConceptScore W3209835134C204030448 @default.
- W3209835134 hasConceptScore W3209835134C2524010 @default.
- W3209835134 hasConceptScore W3209835134C2776135515 @default.
- W3209835134 hasConceptScore W3209835134C33923547 @default.
- W3209835134 hasConceptScore W3209835134C41008148 @default.
- W3209835134 hasConceptScore W3209835134C64543145 @default.
- W3209835134 hasConceptScore W3209835134C74193536 @default.
- W3209835134 hasLocation W32098351341 @default.
- W3209835134 hasOpenAccess W3209835134 @default.
- W3209835134 hasPrimaryLocation W32098351341 @default.
- W3209835134 hasRelatedWork W1016877926 @default.
- W3209835134 hasRelatedWork W1581425933 @default.
- W3209835134 hasRelatedWork W201622579 @default.
- W3209835134 hasRelatedWork W2122308921 @default.
- W3209835134 hasRelatedWork W2258035973 @default.
- W3209835134 hasRelatedWork W2765950489 @default.
- W3209835134 hasRelatedWork W2767274788 @default.
- W3209835134 hasRelatedWork W2801898543 @default.
- W3209835134 hasRelatedWork W2894779738 @default.
- W3209835134 hasRelatedWork W2896989282 @default.
- W3209835134 hasRelatedWork W2903180926 @default.
- W3209835134 hasRelatedWork W2971516670 @default.
- W3209835134 hasRelatedWork W3000254070 @default.