FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3036223397> ?p ?o ?g. }

• W3036223397 abstract "Deep neural network (DNN) training suffers from the significant energy consumption in memory system, and most existing energy reduction techniques for memory system have focused on introducing low precision that is compatible with computing unit (e.g., FP16, FP8). These researches have shown that even in learning the networks with FP16 data precision, it is possible to provide training accuracy as good as FP32, de facto standard of the DNN training. However, our extensive experiments show that we can further reduce the data precision while maintaining the training accuracy of DNNs, which can be obtained by truncating some least significant bits (LSBs) of FP16, named as hard approximation. Nevertheless, the existing hard-ware structures for DNN training cannot efficiently support such low precision. In this work, we propose a novel memory system architecture for GPUs, named as precision-controlled memory system (PCM), which allows for flexible management at the level of hard approximation. PCM provides high DRAM bandwidth by distributing each precision to different channels with as transposed data mapping on DRAM. In addition, PCM supports fine-grained hard approximation in the L1 data cache using software-controlled registers, which can reduce data movement and thereby improve energy saving and system performance. Furthermore, PCM facilitates the reduction of data maintenance energy, which accounts for a considerable portion of memory energy consumption, by controlling refresh period of DRAM. The experimental results show that in training CIFAR-100 dataset on Resnet-20 with precision tuning, PCM achieves energy saving and performance enhancement by 66% and 20%, respectively, without loss of accuracy." @default.
• W3036223397 created "2020-06-25" @default.
• W3036223397 creator A5011045826 @default.
• W3036223397 creator A5018308993 @default.
• W3036223397 creator A5018573573 @default.
• W3036223397 creator A5020778843 @default.
• W3036223397 creator A5021393078 @default.
• W3036223397 creator A5024723558 @default.
• W3036223397 creator A5033091196 @default.
• W3036223397 creator A5041437613 @default.
• W3036223397 creator A5072707532 @default.
• W3036223397 creator A5090060830 @default.
• W3036223397 date "2020-03-01" @default.
• W3036223397 modified "2023-10-05" @default.
• W3036223397 title "PCM: Precision-Controlled Memory System for Energy Efficient Deep Neural Network Training" @default.
• W3036223397 cites W1979527452 @default.
• W3036223397 cites W2049504176 @default.
• W3036223397 cites W2119092821 @default.
• W3036223397 cites W2166918318 @default.
• W3036223397 cites W2194775991 @default.
• W3036223397 cites W2300242332 @default.
• W3036223397 cites W2554302513 @default.
• W3036223397 cites W2725159389 @default.
• W3036223397 cites W2761132374 @default.
• W3036223397 cites W2799050829 @default.
• W3036223397 cites W2800371094 @default.
• W3036223397 cites W2963809228 @default.
• W3036223397 doi "https://doi.org/10.23919/date48585.2020.9116530" @default.
• W3036223397 hasPublicationYear "2020" @default.
• W3036223397 type Work @default.
• W3036223397 sameAs 3036223397 @default.
• W3036223397 citedByCount "5" @default.
• W3036223397 countsByYear W30362233972020 @default.
• W3036223397 countsByYear W30362233972021 @default.
• W3036223397 countsByYear W30362233972022 @default.
• W3036223397 countsByYear W30362233972023 @default.
• W3036223397 crossrefType "proceedings-article" @default.
• W3036223397 hasAuthorship W3036223397A5011045826 @default.
• W3036223397 hasAuthorship W3036223397A5018308993 @default.
• W3036223397 hasAuthorship W3036223397A5018573573 @default.
• W3036223397 hasAuthorship W3036223397A5020778843 @default.
• W3036223397 hasAuthorship W3036223397A5021393078 @default.
• W3036223397 hasAuthorship W3036223397A5024723558 @default.
• W3036223397 hasAuthorship W3036223397A5033091196 @default.
• W3036223397 hasAuthorship W3036223397A5041437613 @default.
• W3036223397 hasAuthorship W3036223397A5072707532 @default.
• W3036223397 hasAuthorship W3036223397A5090060830 @default.
• W3036223397 hasConcept C105795698 @default.
• W3036223397 hasConcept C111335779 @default.
• W3036223397 hasConcept C113775141 @default.
• W3036223397 hasConcept C115537543 @default.
• W3036223397 hasConcept C119599485 @default.
• W3036223397 hasConcept C127413603 @default.
• W3036223397 hasConcept C154945302 @default.
• W3036223397 hasConcept C173608175 @default.
• W3036223397 hasConcept C186370098 @default.
• W3036223397 hasConcept C18903297 @default.
• W3036223397 hasConcept C2524010 @default.
• W3036223397 hasConcept C2742236 @default.
• W3036223397 hasConcept C2780165032 @default.
• W3036223397 hasConcept C33923547 @default.
• W3036223397 hasConcept C41008148 @default.
• W3036223397 hasConcept C50644808 @default.
• W3036223397 hasConcept C68043766 @default.
• W3036223397 hasConcept C7366592 @default.
• W3036223397 hasConcept C86803240 @default.
• W3036223397 hasConcept C9390403 @default.
• W3036223397 hasConceptScore W3036223397C105795698 @default.
• W3036223397 hasConceptScore W3036223397C111335779 @default.
• W3036223397 hasConceptScore W3036223397C113775141 @default.
• W3036223397 hasConceptScore W3036223397C115537543 @default.
• W3036223397 hasConceptScore W3036223397C119599485 @default.
• W3036223397 hasConceptScore W3036223397C127413603 @default.
• W3036223397 hasConceptScore W3036223397C154945302 @default.
• W3036223397 hasConceptScore W3036223397C173608175 @default.
• W3036223397 hasConceptScore W3036223397C186370098 @default.
• W3036223397 hasConceptScore W3036223397C18903297 @default.
• W3036223397 hasConceptScore W3036223397C2524010 @default.
• W3036223397 hasConceptScore W3036223397C2742236 @default.
• W3036223397 hasConceptScore W3036223397C2780165032 @default.
• W3036223397 hasConceptScore W3036223397C33923547 @default.
• W3036223397 hasConceptScore W3036223397C41008148 @default.
• W3036223397 hasConceptScore W3036223397C50644808 @default.
• W3036223397 hasConceptScore W3036223397C68043766 @default.
• W3036223397 hasConceptScore W3036223397C7366592 @default.
• W3036223397 hasConceptScore W3036223397C86803240 @default.
• W3036223397 hasConceptScore W3036223397C9390403 @default.
• W3036223397 hasLocation W30362233971 @default.
• W3036223397 hasOpenAccess W3036223397 @default.
• W3036223397 hasPrimaryLocation W30362233971 @default.
• W3036223397 hasRelatedWork W1973775383 @default.
• W3036223397 hasRelatedWork W2005656440 @default.
• W3036223397 hasRelatedWork W2149765501 @default.
• W3036223397 hasRelatedWork W2152933730 @default.
• W3036223397 hasRelatedWork W2209325467 @default.
• W3036223397 hasRelatedWork W2736888776 @default.
• W3036223397 hasRelatedWork W2885593210 @default.
• W3036223397 hasRelatedWork W3087248942 @default.
• W3036223397 hasRelatedWork W4236285338 @default.
• W3036223397 hasRelatedWork W4242355438 @default.

• next