Matches in SemOpenAlex for { <https://semopenalex.org/work/W3202820162> ?p ?o ?g. }
- W3202820162 endingPage "547" @default.
- W3202820162 startingPage "532" @default.
- W3202820162 abstract "We consider a distributed learning problem in a wireless network, consisting of <inline-formula> <tex-math notation=LaTeX>$N$ </tex-math></inline-formula> distributed edge devices and a parameter server (PS). The objective function is a sum of the edge devices’ local loss functions, who aim to train a shared model by communicating with the PS over multiple access channels (MAC). This problem has attracted a growing interest in distributed sensing systems, and more recently in federated learning, known as over-the-air computation. In this paper, we develop a novel Accelerated Gradient-descent Multiple Access (AGMA) algorithm that uses momentum-based gradient signals over noisy fading MAC to improve the convergence rate as compared to existing methods. Furthermore, AGMA does not require power control or beamforming to cancel the fading effect, which simplifies the implementation complexity. We analyze AGMA theoretically, and establish a finite-sample bound of the error for both convex and strongly convex loss functions with Lipschitz gradient. For the strongly convex case, we show that AGMA approaches the best-known linear convergence rate as the network increases. For the convex case, we show that AGMA significantly improves the sub-linear convergence rate as compared to existing methods. Finally, we present simulation results using real datasets that demonstrate better performance by AGMA." @default.
- W3202820162 created "2021-10-11" @default.
- W3202820162 creator A5002176967 @default.
- W3202820162 creator A5013674401 @default.
- W3202820162 creator A5066085947 @default.
- W3202820162 date "2022-02-01" @default.
- W3202820162 modified "2023-10-02" @default.
- W3202820162 title "Accelerated Gradient Descent Learning Over Multiple Access Fading Channels" @default.
- W3202820162 cites W1978470213 @default.
- W3202820162 cites W1991349164 @default.
- W3202820162 cites W1992208280 @default.
- W3202820162 cites W2007339694 @default.
- W3202820162 cites W2027531092 @default.
- W3202820162 cites W2042255695 @default.
- W3202820162 cites W2085160950 @default.
- W3202820162 cites W2097773828 @default.
- W3202820162 cites W2101453243 @default.
- W3202820162 cites W2119058356 @default.
- W3202820162 cites W2124541940 @default.
- W3202820162 cites W2125993116 @default.
- W3202820162 cites W2127447444 @default.
- W3202820162 cites W2140231372 @default.
- W3202820162 cites W2140473338 @default.
- W3202820162 cites W2140747085 @default.
- W3202820162 cites W2141002933 @default.
- W3202820162 cites W2153757022 @default.
- W3202820162 cites W2154131089 @default.
- W3202820162 cites W2170597157 @default.
- W3202820162 cites W2463103808 @default.
- W3202820162 cites W2542306928 @default.
- W3202820162 cites W2578727492 @default.
- W3202820162 cites W2743681928 @default.
- W3202820162 cites W2892348682 @default.
- W3202820162 cites W2907379776 @default.
- W3202820162 cites W2950044985 @default.
- W3202820162 cites W2969343398 @default.
- W3202820162 cites W2971330699 @default.
- W3202820162 cites W2981138228 @default.
- W3202820162 cites W2991151500 @default.
- W3202820162 cites W2991308345 @default.
- W3202820162 cites W3003959415 @default.
- W3202820162 cites W3006919779 @default.
- W3202820162 cites W3047253814 @default.
- W3202820162 cites W3047465037 @default.
- W3202820162 cites W3081130510 @default.
- W3202820162 cites W3090615085 @default.
- W3202820162 doi "https://doi.org/10.1109/jsac.2021.3118410" @default.
- W3202820162 hasPublicationYear "2022" @default.
- W3202820162 type Work @default.
- W3202820162 sameAs 3202820162 @default.
- W3202820162 citedByCount "10" @default.
- W3202820162 countsByYear W32028201622022 @default.
- W3202820162 countsByYear W32028201622023 @default.
- W3202820162 crossrefType "journal-article" @default.
- W3202820162 hasAuthorship W3202820162A5002176967 @default.
- W3202820162 hasAuthorship W3202820162A5013674401 @default.
- W3202820162 hasAuthorship W3202820162A5066085947 @default.
- W3202820162 hasBestOaLocation W32028201622 @default.
- W3202820162 hasConcept C112680207 @default.
- W3202820162 hasConcept C11413529 @default.
- W3202820162 hasConcept C126255220 @default.
- W3202820162 hasConcept C127162648 @default.
- W3202820162 hasConcept C134306372 @default.
- W3202820162 hasConcept C145446738 @default.
- W3202820162 hasConcept C153258448 @default.
- W3202820162 hasConcept C154945302 @default.
- W3202820162 hasConcept C162307627 @default.
- W3202820162 hasConcept C162324750 @default.
- W3202820162 hasConcept C22324862 @default.
- W3202820162 hasConcept C24590314 @default.
- W3202820162 hasConcept C2524010 @default.
- W3202820162 hasConcept C2777303404 @default.
- W3202820162 hasConcept C31258907 @default.
- W3202820162 hasConcept C33923547 @default.
- W3202820162 hasConcept C41008148 @default.
- W3202820162 hasConcept C50522688 @default.
- W3202820162 hasConcept C50644808 @default.
- W3202820162 hasConcept C54197355 @default.
- W3202820162 hasConcept C57273362 @default.
- W3202820162 hasConcept C57869625 @default.
- W3202820162 hasConcept C76155785 @default.
- W3202820162 hasConcept C81978471 @default.
- W3202820162 hasConceptScore W3202820162C112680207 @default.
- W3202820162 hasConceptScore W3202820162C11413529 @default.
- W3202820162 hasConceptScore W3202820162C126255220 @default.
- W3202820162 hasConceptScore W3202820162C127162648 @default.
- W3202820162 hasConceptScore W3202820162C134306372 @default.
- W3202820162 hasConceptScore W3202820162C145446738 @default.
- W3202820162 hasConceptScore W3202820162C153258448 @default.
- W3202820162 hasConceptScore W3202820162C154945302 @default.
- W3202820162 hasConceptScore W3202820162C162307627 @default.
- W3202820162 hasConceptScore W3202820162C162324750 @default.
- W3202820162 hasConceptScore W3202820162C22324862 @default.
- W3202820162 hasConceptScore W3202820162C24590314 @default.
- W3202820162 hasConceptScore W3202820162C2524010 @default.
- W3202820162 hasConceptScore W3202820162C2777303404 @default.
- W3202820162 hasConceptScore W3202820162C31258907 @default.
- W3202820162 hasConceptScore W3202820162C33923547 @default.