FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2905209374> ?p ?o ?g. }

• W2905209374 abstract "Intelligent Personal Assistants (IPAs) with the capability of natural language processing (NLP) are increasingly popular in today's mobile devices. Recurrent neural networks (RNNs), especially one of their forms - Long-Short Term Memory networks (LSTMs), are becoming the core machine learning technique applied in the NLP-based IPAs. With the continuously improved performance of mobile GPUs, local processing has become a promising solution to the large data transmission and privacy issues induced by the cloud-centric computations of IPAs. However, LSTMs exhibit quite inefficient memory access pattern when executed on mobile GPUs due to the redundant data movements and limited off-chip bandwidth. In this study, we aim to explore the memory friendly LSTM on mobile GPUs by hierarchically reducing the off-chip memory accesses. To address the redundant data movements, we propose inter-cell level optimizations that intelligently parallelize the originally sequentially executed LSTM cells (basic units in RNNs, corresponding to neurons in CNNs) to improve the data locality across cells with negligible accuracy loss. To relax the pressure on limited off-chip memory bandwidth, we propose intra-cell level optimizations that dynamically skip the loads and computations of rows in the weight matrices with trivial contribution to the outputs. We also introduce a light-weighted module to the GPUs architecture for the runtime row skipping in weight matrices. Moreover, our techniques are equipped with thresholds which provide a unique tunning space for performance-accuracy trade-offs directly guided by the user preferences. The experimental results show our optimizations achieves substantial improvements on both performance and power with user-imperceptible accuracy loss. And our optimizations exhibit the strong scalability with the increasing input data set. Our user study also shows that our designed system delivers the excellent user experience." @default.
• W2905209374 created "2018-12-22" @default.
• W2905209374 creator A5010481801 @default.
• W2905209374 creator A5016298390 @default.
• W2905209374 creator A5024508481 @default.
• W2905209374 creator A5037677450 @default.
• W2905209374 creator A5068428294 @default.
• W2905209374 date "2018-10-01" @default.
• W2905209374 modified "2023-09-25" @default.
• W2905209374 title "Towards Memory Friendly Long-Short Term Memory Networks (LSTMs) on Mobile GPUs" @default.
• W2905209374 cites W1632114991 @default.
• W2905209374 cites W1979527452 @default.
• W2905209374 cites W2132774949 @default.
• W2905209374 cites W2163455955 @default.
• W2905209374 cites W2285660444 @default.
• W2905209374 cites W2289252105 @default.
• W2905209374 cites W2346158202 @default.
• W2905209374 cites W2401932869 @default.
• W2905209374 cites W2513554817 @default.
• W2905209374 cites W2527036487 @default.
• W2905209374 cites W2585720638 @default.
• W2905209374 cites W2605258629 @default.
• W2905209374 cites W2613168994 @default.
• W2905209374 cites W2625457103 @default.
• W2905209374 cites W2657126969 @default.
• W2905209374 cites W2728529009 @default.
• W2905209374 cites W2765315405 @default.
• W2905209374 cites W2788838111 @default.
• W2905209374 cites W2794478957 @default.
• W2905209374 cites W2962821792 @default.
• W2905209374 cites W3013279411 @default.
• W2905209374 cites W4240477358 @default.
• W2905209374 cites W4251575795 @default.
• W2905209374 doi "https://doi.org/10.1109/micro.2018.00022" @default.
• W2905209374 hasPublicationYear "2018" @default.
• W2905209374 type Work @default.
• W2905209374 sameAs 2905209374 @default.
• W2905209374 citedByCount "17" @default.
• W2905209374 countsByYear W29052093742019 @default.
• W2905209374 countsByYear W29052093742020 @default.
• W2905209374 countsByYear W29052093742021 @default.
• W2905209374 countsByYear W29052093742022 @default.
• W2905209374 crossrefType "proceedings-article" @default.
• W2905209374 hasAuthorship W2905209374A5010481801 @default.
• W2905209374 hasAuthorship W2905209374A5016298390 @default.
• W2905209374 hasAuthorship W2905209374A5024508481 @default.
• W2905209374 hasAuthorship W2905209374A5037677450 @default.
• W2905209374 hasAuthorship W2905209374A5068428294 @default.
• W2905209374 hasConcept C118524514 @default.
• W2905209374 hasConcept C121332964 @default.
• W2905209374 hasConcept C133488467 @default.
• W2905209374 hasConcept C147168706 @default.
• W2905209374 hasConcept C154945302 @default.
• W2905209374 hasConcept C15744967 @default.
• W2905209374 hasConcept C169760540 @default.
• W2905209374 hasConcept C169900460 @default.
• W2905209374 hasConcept C173608175 @default.
• W2905209374 hasConcept C179226034 @default.
• W2905209374 hasConcept C21963081 @default.
• W2905209374 hasConcept C41008148 @default.
• W2905209374 hasConcept C50644808 @default.
• W2905209374 hasConcept C61797465 @default.
• W2905209374 hasConcept C62520636 @default.
• W2905209374 hasConceptScore W2905209374C118524514 @default.
• W2905209374 hasConceptScore W2905209374C121332964 @default.
• W2905209374 hasConceptScore W2905209374C133488467 @default.
• W2905209374 hasConceptScore W2905209374C147168706 @default.
• W2905209374 hasConceptScore W2905209374C154945302 @default.
• W2905209374 hasConceptScore W2905209374C15744967 @default.
• W2905209374 hasConceptScore W2905209374C169760540 @default.
• W2905209374 hasConceptScore W2905209374C169900460 @default.
• W2905209374 hasConceptScore W2905209374C173608175 @default.
• W2905209374 hasConceptScore W2905209374C179226034 @default.
• W2905209374 hasConceptScore W2905209374C21963081 @default.
• W2905209374 hasConceptScore W2905209374C41008148 @default.
• W2905209374 hasConceptScore W2905209374C50644808 @default.
• W2905209374 hasConceptScore W2905209374C61797465 @default.
• W2905209374 hasConceptScore W2905209374C62520636 @default.
• W2905209374 hasLocation W29052093741 @default.
• W2905209374 hasOpenAccess W2905209374 @default.
• W2905209374 hasPrimaryLocation W29052093741 @default.
• W2905209374 hasRelatedWork W2034384303 @default.
• W2905209374 hasRelatedWork W2904581126 @default.
• W2905209374 hasRelatedWork W2917883546 @default.
• W2905209374 hasRelatedWork W2954151159 @default.
• W2905209374 hasRelatedWork W2979437663 @default.
• W2905209374 hasRelatedWork W3016404724 @default.
• W2905209374 hasRelatedWork W3155820633 @default.
• W2905209374 hasRelatedWork W3163603179 @default.
• W2905209374 hasRelatedWork W4225519494 @default.
• W2905209374 hasRelatedWork W4255499636 @default.
• W2905209374 isParatext "false" @default.
• W2905209374 isRetracted "false" @default.
• W2905209374 magId "2905209374" @default.
• W2905209374 workType "article" @default.