Matches in SemOpenAlex for { <https://semopenalex.org/work/W3102122007> ?p ?o ?g. }
- W3102122007 abstract "In recent years, with the popularization of deep learning frameworks and large datasets, researchers have started parallelizing their models in order to train faster. This is crucially important, because they typically explore many hyperparameters in order to find the best ones for their applications. This process is time consuming and, consequently, speeding up training improves productivity. One approach to parallelize deep learning models followed by many researchers is based on weak scaling. The minibatches increase in size as new GPUs are added to the system. In addition, new learning rates schedules have been proposed to fix optimization issues that occur with large minibatch sizes. In this paper, however, we show that the recommendations provided by recent work do not apply to models that lack large datasets. In fact, we argument in favor of using strong scaling for achieving reliable performance in such cases. We evaluated our approach with up to 32 GPUs and show that weak scaling not only does not have the same accuracy as the sequential model, it also fails to converge most of time. Meanwhile, strong scaling has good scalability while having exactly the same accuracy of a sequential implementation." @default.
- W3102122007 created "2020-11-23" @default.
- W3102122007 creator A5020569547 @default.
- W3102122007 creator A5051251449 @default.
- W3102122007 creator A5064637719 @default.
- W3102122007 creator A5086325223 @default.
- W3102122007 date "2018-09-01" @default.
- W3102122007 modified "2023-10-03" @default.
- W3102122007 title "An Argument in Favor of Strong Scaling for Deep Neural Networks with Small Datasets" @default.
- W3102122007 cites W1442374986 @default.
- W3102122007 cites W1522301498 @default.
- W3102122007 cites W1757796397 @default.
- W3102122007 cites W1979566015 @default.
- W3102122007 cites W1988115241 @default.
- W3102122007 cites W2041823554 @default.
- W3102122007 cites W2051567790 @default.
- W3102122007 cites W2071925767 @default.
- W3102122007 cites W2145339207 @default.
- W3102122007 cites W2163605009 @default.
- W3102122007 cites W2194775991 @default.
- W3102122007 cites W2402144811 @default.
- W3102122007 cites W2533800772 @default.
- W3102122007 cites W2604744755 @default.
- W3102122007 cites W2613718673 @default.
- W3102122007 cites W2622263826 @default.
- W3102122007 cites W2749988060 @default.
- W3102122007 cites W2766164908 @default.
- W3102122007 cites W2787998955 @default.
- W3102122007 cites W2875583934 @default.
- W3102122007 cites W2892209285 @default.
- W3102122007 cites W2919115771 @default.
- W3102122007 cites W2949650786 @default.
- W3102122007 cites W2963702144 @default.
- W3102122007 cites W2964121744 @default.
- W3102122007 cites W2964308564 @default.
- W3102122007 doi "https://doi.org/10.1109/cahpc.2018.8645881" @default.
- W3102122007 hasPublicationYear "2018" @default.
- W3102122007 type Work @default.
- W3102122007 sameAs 3102122007 @default.
- W3102122007 citedByCount "1" @default.
- W3102122007 countsByYear W31021220072020 @default.
- W3102122007 crossrefType "proceedings-article" @default.
- W3102122007 hasAuthorship W3102122007A5020569547 @default.
- W3102122007 hasAuthorship W3102122007A5051251449 @default.
- W3102122007 hasAuthorship W3102122007A5064637719 @default.
- W3102122007 hasAuthorship W3102122007A5086325223 @default.
- W3102122007 hasBestOaLocation W31021220072 @default.
- W3102122007 hasConcept C10138342 @default.
- W3102122007 hasConcept C108583219 @default.
- W3102122007 hasConcept C119857082 @default.
- W3102122007 hasConcept C154945302 @default.
- W3102122007 hasConcept C162324750 @default.
- W3102122007 hasConcept C182306322 @default.
- W3102122007 hasConcept C185592680 @default.
- W3102122007 hasConcept C199360897 @default.
- W3102122007 hasConcept C2524010 @default.
- W3102122007 hasConcept C2984842247 @default.
- W3102122007 hasConcept C33923547 @default.
- W3102122007 hasConcept C41008148 @default.
- W3102122007 hasConcept C48044578 @default.
- W3102122007 hasConcept C50644808 @default.
- W3102122007 hasConcept C55493867 @default.
- W3102122007 hasConcept C77088390 @default.
- W3102122007 hasConcept C8642999 @default.
- W3102122007 hasConcept C98045186 @default.
- W3102122007 hasConcept C98184364 @default.
- W3102122007 hasConcept C99844830 @default.
- W3102122007 hasConceptScore W3102122007C10138342 @default.
- W3102122007 hasConceptScore W3102122007C108583219 @default.
- W3102122007 hasConceptScore W3102122007C119857082 @default.
- W3102122007 hasConceptScore W3102122007C154945302 @default.
- W3102122007 hasConceptScore W3102122007C162324750 @default.
- W3102122007 hasConceptScore W3102122007C182306322 @default.
- W3102122007 hasConceptScore W3102122007C185592680 @default.
- W3102122007 hasConceptScore W3102122007C199360897 @default.
- W3102122007 hasConceptScore W3102122007C2524010 @default.
- W3102122007 hasConceptScore W3102122007C2984842247 @default.
- W3102122007 hasConceptScore W3102122007C33923547 @default.
- W3102122007 hasConceptScore W3102122007C41008148 @default.
- W3102122007 hasConceptScore W3102122007C48044578 @default.
- W3102122007 hasConceptScore W3102122007C50644808 @default.
- W3102122007 hasConceptScore W3102122007C55493867 @default.
- W3102122007 hasConceptScore W3102122007C77088390 @default.
- W3102122007 hasConceptScore W3102122007C8642999 @default.
- W3102122007 hasConceptScore W3102122007C98045186 @default.
- W3102122007 hasConceptScore W3102122007C98184364 @default.
- W3102122007 hasConceptScore W3102122007C99844830 @default.
- W3102122007 hasLocation W31021220071 @default.
- W3102122007 hasLocation W31021220072 @default.
- W3102122007 hasOpenAccess W3102122007 @default.
- W3102122007 hasPrimaryLocation W31021220071 @default.
- W3102122007 hasRelatedWork W2576264401 @default.
- W3102122007 hasRelatedWork W2978367927 @default.
- W3102122007 hasRelatedWork W3047644063 @default.
- W3102122007 hasRelatedWork W3102122007 @default.
- W3102122007 hasRelatedWork W3121832479 @default.
- W3102122007 hasRelatedWork W3174422331 @default.
- W3102122007 hasRelatedWork W3209881118 @default.
- W3102122007 hasRelatedWork W4210794429 @default.
- W3102122007 hasRelatedWork W4280644903 @default.