Matches in SemOpenAlex for { <https://semopenalex.org/work/W4226137788> ?p ?o ?g. }
Showing items 1 to 95 of
95
with 100 items per page.
- W4226137788 abstract "Graph Convolutional Networks (GCNs) have emerged as the state-of-the-art graph learning model. However, it can be notoriously challenging to inference GCNs over large graph datasets, limiting their application to large real-world graphs and hindering the exploration of deeper and more sophisticated GCN graphs. This is because real-world graphs can be extremely large and sparse. Furthermore, the node degree of GCNs tends to follow the power-law distribution and therefore have highly irregular adjacency matrices, resulting in prohibitive inefficiencies in both data processing and movement and thus substantially limiting the achievable GCN acceleration efficiency. To this end, this paper proposes a GCN algorithm and accelerator Co-Design framework dubbed GCoD which can largely alleviate the aforementioned GCN irregularity and boost GCNs’ inference efficiency. Specifically, on the algorithm level, GCoD integrates a split and conquer GCN training strategy that polarizes the graphs to be either denser or sparser in local neighborhoods without compromising the model accuracy, resulting in graph adjacency matrices that (mostly) have merely two levels of workload and enjoys largely enhanced regularity and thus ease of acceleration. On the hardware level, we further develop a dedicated two-pronged accelerator with a separated engine to process each of the aforementioned denser and sparser workloads, further boosting the overall utilization and acceleration efficiency. Extensive experiments and ablation studies validate that our GCoD consistently reduces the number of off-chip accesses, leading to speedups 15286×, 294×, 7.8×, and 2.5× as compared to CPUs, GPUs, and prior-art GCN accelerators including HyGCN and AWB-GCN, respectively, while maintaining or even improving the task accuracy. Additionally, we visualize GCoD trained graph adjacency matrices for a better understanding of its advantages." @default.
- W4226137788 created "2022-05-05" @default.
- W4226137788 creator A5019582323 @default.
- W4226137788 creator A5059605240 @default.
- W4226137788 creator A5075140744 @default.
- W4226137788 creator A5078443672 @default.
- W4226137788 creator A5080793133 @default.
- W4226137788 date "2022-04-01" @default.
- W4226137788 modified "2023-10-16" @default.
- W4226137788 title "GCoD: Graph Convolutional Network Acceleration via Dedicated Algorithm and Accelerator Co-Design" @default.
- W4226137788 cites W2070232376 @default.
- W4226137788 cites W2153959628 @default.
- W4226137788 cites W2466675884 @default.
- W4226137788 cites W2484446135 @default.
- W4226137788 cites W2539871928 @default.
- W4226137788 cites W2759777444 @default.
- W4226137788 cites W2766362889 @default.
- W4226137788 cites W2860338957 @default.
- W4226137788 cites W2898565872 @default.
- W4226137788 cites W2964571482 @default.
- W4226137788 cites W2996835428 @default.
- W4226137788 cites W3017228913 @default.
- W4226137788 cites W3021747854 @default.
- W4226137788 cites W3091862797 @default.
- W4226137788 cites W3105753905 @default.
- W4226137788 cites W3106125969 @default.
- W4226137788 cites W3111579839 @default.
- W4226137788 cites W3137187143 @default.
- W4226137788 cites W3200832253 @default.
- W4226137788 doi "https://doi.org/10.1109/hpca53966.2022.00041" @default.
- W4226137788 hasPublicationYear "2022" @default.
- W4226137788 type Work @default.
- W4226137788 citedByCount "9" @default.
- W4226137788 countsByYear W42261377882022 @default.
- W4226137788 countsByYear W42261377882023 @default.
- W4226137788 crossrefType "proceedings-article" @default.
- W4226137788 hasAuthorship W4226137788A5019582323 @default.
- W4226137788 hasAuthorship W4226137788A5059605240 @default.
- W4226137788 hasAuthorship W4226137788A5075140744 @default.
- W4226137788 hasAuthorship W4226137788A5078443672 @default.
- W4226137788 hasAuthorship W4226137788A5080793133 @default.
- W4226137788 hasBestOaLocation W42261377882 @default.
- W4226137788 hasConcept C113775141 @default.
- W4226137788 hasConcept C11413529 @default.
- W4226137788 hasConcept C117896860 @default.
- W4226137788 hasConcept C121332964 @default.
- W4226137788 hasConcept C127413603 @default.
- W4226137788 hasConcept C13251829 @default.
- W4226137788 hasConcept C132525143 @default.
- W4226137788 hasConcept C154945302 @default.
- W4226137788 hasConcept C173608175 @default.
- W4226137788 hasConcept C180356752 @default.
- W4226137788 hasConcept C188198153 @default.
- W4226137788 hasConcept C203776342 @default.
- W4226137788 hasConcept C2776214188 @default.
- W4226137788 hasConcept C41008148 @default.
- W4226137788 hasConcept C43517604 @default.
- W4226137788 hasConcept C74650414 @default.
- W4226137788 hasConcept C78519656 @default.
- W4226137788 hasConcept C80444323 @default.
- W4226137788 hasConceptScore W4226137788C113775141 @default.
- W4226137788 hasConceptScore W4226137788C11413529 @default.
- W4226137788 hasConceptScore W4226137788C117896860 @default.
- W4226137788 hasConceptScore W4226137788C121332964 @default.
- W4226137788 hasConceptScore W4226137788C127413603 @default.
- W4226137788 hasConceptScore W4226137788C13251829 @default.
- W4226137788 hasConceptScore W4226137788C132525143 @default.
- W4226137788 hasConceptScore W4226137788C154945302 @default.
- W4226137788 hasConceptScore W4226137788C173608175 @default.
- W4226137788 hasConceptScore W4226137788C180356752 @default.
- W4226137788 hasConceptScore W4226137788C188198153 @default.
- W4226137788 hasConceptScore W4226137788C203776342 @default.
- W4226137788 hasConceptScore W4226137788C2776214188 @default.
- W4226137788 hasConceptScore W4226137788C41008148 @default.
- W4226137788 hasConceptScore W4226137788C43517604 @default.
- W4226137788 hasConceptScore W4226137788C74650414 @default.
- W4226137788 hasConceptScore W4226137788C78519656 @default.
- W4226137788 hasConceptScore W4226137788C80444323 @default.
- W4226137788 hasLocation W42261377881 @default.
- W4226137788 hasLocation W42261377882 @default.
- W4226137788 hasOpenAccess W4226137788 @default.
- W4226137788 hasPrimaryLocation W42261377881 @default.
- W4226137788 hasRelatedWork W2014256025 @default.
- W4226137788 hasRelatedWork W2355108509 @default.
- W4226137788 hasRelatedWork W2470825435 @default.
- W4226137788 hasRelatedWork W2951515642 @default.
- W4226137788 hasRelatedWork W3034334914 @default.
- W4226137788 hasRelatedWork W3168657718 @default.
- W4226137788 hasRelatedWork W3202522988 @default.
- W4226137788 hasRelatedWork W4226315598 @default.
- W4226137788 hasRelatedWork W4304700924 @default.
- W4226137788 hasRelatedWork W4320342417 @default.
- W4226137788 isParatext "false" @default.
- W4226137788 isRetracted "false" @default.
- W4226137788 workType "article" @default.