Matches in SemOpenAlex for { <https://semopenalex.org/work/W3038531770> ?p ?o ?g. }
- W3038531770 endingPage "3708" @default.
- W3038531770 startingPage "3693" @default.
- W3038531770 abstract "Stochastic gradient descent is a canonical tool for addressing stochastic optimization problems, which forms the bedrock of modern machine learning. In this work, we seek to balance the fact that attenuating step-size is required for exact convergence with the fact that constant step-size learns faster to a limiting error. To do so, rather than fixing the mini-batch and the step-size at the outset, we propose a strategy to allow parameters evolving adaptively. Specifically, the batch-size is set to be a piecewise-constant increasing sequence where the increase occurs when a suitable error criterion is satisfied. Moreover, the step-size is selected as that which yields the fastest convergence. The overall algorithm, two scale adaptive (TSA) scheme, is developed for both convex and non-convex problems. It inherits the exact convergence and more importantly, the optimal error decreasing rate and an overall computation reduction are achieved. Furthermore, we extended the TSA method to the generalized adaptive batching framework, which is a generic methodology modular to any stochastic algorithms pursuing a trade-off between convergence rates and stochastic variance. We evaluate the TSA method on the image classification problem on MNIST and CIFAR-10 datasets compared with standard SGD methods and existing adaptive batch-size methods, to corroborate theoretical findings." @default.
- W3038531770 created "2020-07-10" @default.
- W3038531770 creator A5025896653 @default.
- W3038531770 creator A5078862959 @default.
- W3038531770 creator A5088013951 @default.
- W3038531770 date "2022-01-01" @default.
- W3038531770 modified "2023-09-25" @default.
- W3038531770 title "Balancing Rates and Variance via Adaptive Batch-Size for Stochastic Optimization Problems" @default.
- W3038531770 cites W1507126399 @default.
- W3038531770 cites W1774344329 @default.
- W3038531770 cites W1994616650 @default.
- W3038531770 cites W2001905757 @default.
- W3038531770 cites W2004544971 @default.
- W3038531770 cites W2020909452 @default.
- W3038531770 cites W2030161963 @default.
- W3038531770 cites W2034996255 @default.
- W3038531770 cites W2038084402 @default.
- W3038531770 cites W2045551057 @default.
- W3038531770 cites W2052725501 @default.
- W3038531770 cites W2061570747 @default.
- W3038531770 cites W2072566913 @default.
- W3038531770 cites W2086161653 @default.
- W3038531770 cites W2095984592 @default.
- W3038531770 cites W2103628399 @default.
- W3038531770 cites W2118439011 @default.
- W3038531770 cites W2124541940 @default.
- W3038531770 cites W2132749914 @default.
- W3038531770 cites W2294586855 @default.
- W3038531770 cites W2405727896 @default.
- W3038531770 cites W2556219822 @default.
- W3038531770 cites W2907379776 @default.
- W3038531770 cites W2963433607 @default.
- W3038531770 cites W2963470657 @default.
- W3038531770 cites W2963941964 @default.
- W3038531770 cites W2964026281 @default.
- W3038531770 cites W3015618519 @default.
- W3038531770 doi "https://doi.org/10.1109/tsp.2022.3186526" @default.
- W3038531770 hasPublicationYear "2022" @default.
- W3038531770 type Work @default.
- W3038531770 sameAs 3038531770 @default.
- W3038531770 citedByCount "1" @default.
- W3038531770 countsByYear W30385317702023 @default.
- W3038531770 crossrefType "journal-article" @default.
- W3038531770 hasAuthorship W3038531770A5025896653 @default.
- W3038531770 hasAuthorship W3038531770A5078862959 @default.
- W3038531770 hasAuthorship W3038531770A5088013951 @default.
- W3038531770 hasBestOaLocation W30385317702 @default.
- W3038531770 hasConcept C105795698 @default.
- W3038531770 hasConcept C108583219 @default.
- W3038531770 hasConcept C11413529 @default.
- W3038531770 hasConcept C126255220 @default.
- W3038531770 hasConcept C134306372 @default.
- W3038531770 hasConcept C154945302 @default.
- W3038531770 hasConcept C162324750 @default.
- W3038531770 hasConcept C164660894 @default.
- W3038531770 hasConcept C190502265 @default.
- W3038531770 hasConcept C194387892 @default.
- W3038531770 hasConcept C19499675 @default.
- W3038531770 hasConcept C206688291 @default.
- W3038531770 hasConcept C26517878 @default.
- W3038531770 hasConcept C2777303404 @default.
- W3038531770 hasConcept C33923547 @default.
- W3038531770 hasConcept C38652104 @default.
- W3038531770 hasConcept C41008148 @default.
- W3038531770 hasConcept C50522688 @default.
- W3038531770 hasConcept C50644808 @default.
- W3038531770 hasConcept C55479107 @default.
- W3038531770 hasConcept C57869625 @default.
- W3038531770 hasConcept C62644790 @default.
- W3038531770 hasConceptScore W3038531770C105795698 @default.
- W3038531770 hasConceptScore W3038531770C108583219 @default.
- W3038531770 hasConceptScore W3038531770C11413529 @default.
- W3038531770 hasConceptScore W3038531770C126255220 @default.
- W3038531770 hasConceptScore W3038531770C134306372 @default.
- W3038531770 hasConceptScore W3038531770C154945302 @default.
- W3038531770 hasConceptScore W3038531770C162324750 @default.
- W3038531770 hasConceptScore W3038531770C164660894 @default.
- W3038531770 hasConceptScore W3038531770C190502265 @default.
- W3038531770 hasConceptScore W3038531770C194387892 @default.
- W3038531770 hasConceptScore W3038531770C19499675 @default.
- W3038531770 hasConceptScore W3038531770C206688291 @default.
- W3038531770 hasConceptScore W3038531770C26517878 @default.
- W3038531770 hasConceptScore W3038531770C2777303404 @default.
- W3038531770 hasConceptScore W3038531770C33923547 @default.
- W3038531770 hasConceptScore W3038531770C38652104 @default.
- W3038531770 hasConceptScore W3038531770C41008148 @default.
- W3038531770 hasConceptScore W3038531770C50522688 @default.
- W3038531770 hasConceptScore W3038531770C50644808 @default.
- W3038531770 hasConceptScore W3038531770C55479107 @default.
- W3038531770 hasConceptScore W3038531770C57869625 @default.
- W3038531770 hasConceptScore W3038531770C62644790 @default.
- W3038531770 hasLocation W30385317701 @default.
- W3038531770 hasLocation W30385317702 @default.
- W3038531770 hasOpenAccess W3038531770 @default.
- W3038531770 hasPrimaryLocation W30385317701 @default.
- W3038531770 hasRelatedWork W2075181955 @default.
- W3038531770 hasRelatedWork W2804251232 @default.
- W3038531770 hasRelatedWork W2952221746 @default.