Please use this identifier to cite or link to this item:
https://doi.org/10.1109/TC.2020.3001033
DC Field | Value | |
---|---|---|
dc.title | Accelerating Generative Neural Networks on Unmodified Deep Learning Processors-A Software Approach | |
dc.contributor.author | Xu, Dawen | |
dc.contributor.author | Liu, Cheng | |
dc.contributor.author | Wang, Ying | |
dc.contributor.author | Tu, Kaijie | |
dc.contributor.author | He, Bingsheng | |
dc.contributor.author | Zhang, Lei | |
dc.date.accessioned | 2022-02-15T04:01:08Z | |
dc.date.available | 2022-02-15T04:01:08Z | |
dc.date.issued | 2020-01-08 | |
dc.identifier.citation | Xu, Dawen, Liu, Cheng, Wang, Ying, Tu, Kaijie, He, Bingsheng, Zhang, Lei (2020-01-08). Accelerating Generative Neural Networks on Unmodified Deep Learning Processors-A Software Approach. IEEE TRANSACTIONS ON COMPUTERS 69 (8) : 1172-1184. ScholarBank@NUS Repository. https://doi.org/10.1109/TC.2020.3001033 | |
dc.identifier.issn | 0018-9340 | |
dc.identifier.issn | 1557-9956 | |
dc.identifier.uri | https://scholarbank.nus.edu.sg/handle/10635/215371 | |
dc.description.abstract | Generative neural network is a new category of neural networks and it has been widely utilized in many applications such as content generation, unsupervised learning, segmentation, and pose estimation. It typically involves massive computing-intensive deconvolution operations that cannot be fitted to conventional neural network processors directly. However, prior works mainly investigated specialized hardware architectures through intensive hardware modifications to the existing deep learning processors to accelerate deconvolution together with the convolution. In contrast, this article proposes a novel deconvolution implementation with a software approach and enables fast and efficient deconvolution execution on the existing deep learning processors. Our proposed method reorganizes the computation of deconvolution and allows the deep learning processors to treat it as the standard convolution by splitting the original deconvolution filters into multiple small filters. Compared to prior acceleration schemes, the implemented acceleration scheme achieves 2.4× -4.3× performance speedup and reduces the energy consumption by 27.7 -54.5 percent on a set of realistic benchmarks. In addition, we have also applied the deconvolution computing approach to the off-the-shelf commodity deep learning processors. The performance of deconvolution also exhibits significant performance speedup over prior deconvolution implementations. | |
dc.language.iso | en | |
dc.publisher | IEEE COMPUTER SOC | |
dc.source | Elements | |
dc.subject | Science & Technology | |
dc.subject | Technology | |
dc.subject | Computer Science, Hardware & Architecture | |
dc.subject | Engineering, Electrical & Electronic | |
dc.subject | Computer Science | |
dc.subject | Engineering | |
dc.subject | Deconvolution | |
dc.subject | Program processors | |
dc.subject | Neural networks | |
dc.subject | Convolution | |
dc.subject | Computer architecture | |
dc.subject | Hardware | |
dc.subject | Acceleration | |
dc.subject | Generative neural network | |
dc.subject | deconvolution accelerator | |
dc.subject | split deconvolution | |
dc.type | Article | |
dc.date.updated | 2022-02-14T23:40:59Z | |
dc.contributor.department | DEAN'S OFFICE (SCHOOL OF COMPUTING) | |
dc.description.doi | 10.1109/TC.2020.3001033 | |
dc.description.sourcetitle | IEEE TRANSACTIONS ON COMPUTERS | |
dc.description.volume | 69 | |
dc.description.issue | 8 | |
dc.description.page | 1172-1184 | |
dc.published.state | Published | |
Appears in Collections: | Staff Publications Elements |
Show simple item record
Files in This Item:
File | Description | Size | Format | Access Settings | Version | |
---|---|---|---|---|---|---|
1907.01773v3.pdf | 4.44 MB | Adobe PDF | OPEN | Post-print | View/Download |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.