Flexible Low Power CNN Accelerator for Edge Computing with Weight Tuning
To support various edge applications, a neural network accelerator needs to achieve high flexibility and classification accuracy within a limited power budget. This paper proposes a weight tuning algorithm to improve the energy efficiency by lowering the switching activity. A flexible and runtime-re...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Published: |
Institute of Electrical and Electronics Engineers (IEEE)
2020
|
| Online Access: | https://hdl.handle.net/1721.1/125824 |
| _version_ | 1811082996401307648 |
|---|---|
| author | Wang, Miaorong Chandrakasan, Anantha P |
| author2 | Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science |
| author_facet | Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science Wang, Miaorong Chandrakasan, Anantha P |
| author_sort | Wang, Miaorong |
| collection | MIT |
| description | To support various edge applications, a neural network accelerator needs to achieve high flexibility and classification accuracy within a limited power budget. This paper proposes a weight tuning algorithm to improve the energy efficiency by lowering the switching activity. A flexible and runtime-reconfigurable CNN accelerator is co-designed with the algorithm and demonstrated with a feature extraction processor on an FPGA. The system is fully self-contained for small CNNs and speech keyword spotting is shown as an example. A fully integrated custom ASIC is also being fabricated for this system. Based on post place-and-route simulation of the ASIC, the weight tuning algorithm reduces the energy consumption of weight delivery and computation by 1.70x and 1.20x respectively with little loss in accuracy. |
| first_indexed | 2024-09-23T12:17:07Z |
| format | Article |
| id | mit-1721.1/125824 |
| institution | Massachusetts Institute of Technology |
| last_indexed | 2024-09-23T12:17:07Z |
| publishDate | 2020 |
| publisher | Institute of Electrical and Electronics Engineers (IEEE) |
| record_format | dspace |
| spelling | mit-1721.1/1258242022-10-01T08:52:56Z Flexible Low Power CNN Accelerator for Edge Computing with Weight Tuning Wang, Miaorong Chandrakasan, Anantha P Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science To support various edge applications, a neural network accelerator needs to achieve high flexibility and classification accuracy within a limited power budget. This paper proposes a weight tuning algorithm to improve the energy efficiency by lowering the switching activity. A flexible and runtime-reconfigurable CNN accelerator is co-designed with the algorithm and demonstrated with a feature extraction processor on an FPGA. The system is fully self-contained for small CNNs and speech keyword spotting is shown as an example. A fully integrated custom ASIC is also being fabricated for this system. Based on post place-and-route simulation of the ASIC, the weight tuning algorithm reduces the energy consumption of weight delivery and computation by 1.70x and 1.20x respectively with little loss in accuracy. 2020-06-16T19:18:41Z 2020-06-16T19:18:41Z 2020-04 2019-11 Article http://purl.org/eprint/type/ConferencePaper 9781728151069 https://hdl.handle.net/1721.1/125824 Wang, Miaorong and Anantha P. Chandrakasan. "Flexible Low Power CNN Accelerator for Edge Computing with Weight Tuning." IEEE Asian Solid-State Circuits Conference (A-SSCC), November 2019, Macau, Macao, Institute of Electrical and Electronics Engineers (IEEE), April 2020 http://dx.doi.org/10.1109/a-sscc47793.2019.9056941 IEEE Asian Solid-State Circuits Conference (A-SSCC) Creative Commons Attribution-Noncommercial-Share Alike http://creativecommons.org/licenses/by-nc-sa/4.0/ application/pdf Institute of Electrical and Electronics Engineers (IEEE) Utsav Banerjee |
| spellingShingle | Wang, Miaorong Chandrakasan, Anantha P Flexible Low Power CNN Accelerator for Edge Computing with Weight Tuning |
| title | Flexible Low Power CNN Accelerator for Edge Computing with Weight Tuning |
| title_full | Flexible Low Power CNN Accelerator for Edge Computing with Weight Tuning |
| title_fullStr | Flexible Low Power CNN Accelerator for Edge Computing with Weight Tuning |
| title_full_unstemmed | Flexible Low Power CNN Accelerator for Edge Computing with Weight Tuning |
| title_short | Flexible Low Power CNN Accelerator for Edge Computing with Weight Tuning |
| title_sort | flexible low power cnn accelerator for edge computing with weight tuning |
| url | https://hdl.handle.net/1721.1/125824 |
| work_keys_str_mv | AT wangmiaorong flexiblelowpowercnnacceleratorforedgecomputingwithweighttuning AT chandrakasanananthap flexiblelowpowercnnacceleratorforedgecomputingwithweighttuning |