Matches in SemOpenAlex for { <https://semopenalex.org/work/W4386942593> ?p ?o ?g. }
Showing items 1 to 65 of
65
with 100 items per page.
- W4386942593 abstract "This work presents a new method for enhancing communication efficiency in stochastic Federated Learning that trains over-parameterized random networks. In this setting, a binary mask is optimized instead of the model weights, which are kept fixed. The mask characterizes a sparse sub-network that is able to generalize as good as a smaller target network. Importantly, sparse binary masks are exchanged rather than the floating point weights in traditional federated learning, reducing communication cost to at most 1 bit per parameter. We show that previous state of the art stochastic methods fail to find the sparse networks that can reduce the communication and storage overhead using consistent loss objectives. To address this, we propose adding a regularization term to local objectives that encourages sparser solutions by eliminating redundant features across sub-networks. Extensive experiments demonstrate significant improvements in communication and memory efficiency of up to five magnitudes compared to the literature, with minimal performance degradation in validation accuracy in some instances." @default.
- W4386942593 created "2023-09-22" @default.
- W4386942593 creator A5008606893 @default.
- W4386942593 creator A5021728919 @default.
- W4386942593 creator A5027541114 @default.
- W4386942593 creator A5037504185 @default.
- W4386942593 creator A5060555620 @default.
- W4386942593 date "2023-09-19" @default.
- W4386942593 modified "2023-09-28" @default.
- W4386942593 title "Sparser Random Networks Exist: Enforcing Communication-Efficient Federated Learning via Regularization" @default.
- W4386942593 doi "https://doi.org/10.48550/arxiv.2309.10834" @default.
- W4386942593 hasPublicationYear "2023" @default.
- W4386942593 type Work @default.
- W4386942593 citedByCount "0" @default.
- W4386942593 crossrefType "posted-content" @default.
- W4386942593 hasAuthorship W4386942593A5008606893 @default.
- W4386942593 hasAuthorship W4386942593A5021728919 @default.
- W4386942593 hasAuthorship W4386942593A5027541114 @default.
- W4386942593 hasAuthorship W4386942593A5037504185 @default.
- W4386942593 hasAuthorship W4386942593A5060555620 @default.
- W4386942593 hasBestOaLocation W43869425931 @default.
- W4386942593 hasConcept C111919701 @default.
- W4386942593 hasConcept C11413529 @default.
- W4386942593 hasConcept C120314980 @default.
- W4386942593 hasConcept C124066611 @default.
- W4386942593 hasConcept C154945302 @default.
- W4386942593 hasConcept C165464430 @default.
- W4386942593 hasConcept C2776135515 @default.
- W4386942593 hasConcept C2779960059 @default.
- W4386942593 hasConcept C2988886741 @default.
- W4386942593 hasConcept C33923547 @default.
- W4386942593 hasConcept C41008148 @default.
- W4386942593 hasConcept C48372109 @default.
- W4386942593 hasConcept C80444323 @default.
- W4386942593 hasConcept C94375191 @default.
- W4386942593 hasConceptScore W4386942593C111919701 @default.
- W4386942593 hasConceptScore W4386942593C11413529 @default.
- W4386942593 hasConceptScore W4386942593C120314980 @default.
- W4386942593 hasConceptScore W4386942593C124066611 @default.
- W4386942593 hasConceptScore W4386942593C154945302 @default.
- W4386942593 hasConceptScore W4386942593C165464430 @default.
- W4386942593 hasConceptScore W4386942593C2776135515 @default.
- W4386942593 hasConceptScore W4386942593C2779960059 @default.
- W4386942593 hasConceptScore W4386942593C2988886741 @default.
- W4386942593 hasConceptScore W4386942593C33923547 @default.
- W4386942593 hasConceptScore W4386942593C41008148 @default.
- W4386942593 hasConceptScore W4386942593C48372109 @default.
- W4386942593 hasConceptScore W4386942593C80444323 @default.
- W4386942593 hasConceptScore W4386942593C94375191 @default.
- W4386942593 hasLocation W43869425931 @default.
- W4386942593 hasOpenAccess W4386942593 @default.
- W4386942593 hasPrimaryLocation W43869425931 @default.
- W4386942593 hasRelatedWork W1596201972 @default.
- W4386942593 hasRelatedWork W1973679146 @default.
- W4386942593 hasRelatedWork W2028024605 @default.
- W4386942593 hasRelatedWork W2092071486 @default.
- W4386942593 hasRelatedWork W2160425906 @default.
- W4386942593 hasRelatedWork W2357085366 @default.
- W4386942593 hasRelatedWork W2391167130 @default.
- W4386942593 hasRelatedWork W2767525681 @default.
- W4386942593 hasRelatedWork W4283067488 @default.
- W4386942593 hasRelatedWork W2460246254 @default.
- W4386942593 isParatext "false" @default.
- W4386942593 isRetracted "false" @default.
- W4386942593 workType "article" @default.