StrideHD: A Binary Hyperdimensional Computing System Utilizing Window Striding for Image Classification
Hyper-Dimensional (HD) computing is a brain-inspired learning approach for efficient and fast learning on today’s embedded devices. HDC first encodes all data points to high-dimensional vectors called hypervectors and then efficiently performs the classification task using a well-defined...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2024-01-01
|
| Series: | IEEE Open Journal of Circuits and Systems |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10530353/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832592874601971712 |
|---|---|
| author | Dehua Liang Jun Shiomi Noriyuki Miura Hiromitsu Awano |
| author_facet | Dehua Liang Jun Shiomi Noriyuki Miura Hiromitsu Awano |
| author_sort | Dehua Liang |
| collection | DOAJ |
| description | Hyper-Dimensional (HD) computing is a brain-inspired learning approach for efficient and fast learning on today’s embedded devices. HDC first encodes all data points to high-dimensional vectors called hypervectors and then efficiently performs the classification task using a well-defined set of operations. Although HDC achieved reasonable performances in several practical tasks, it comes with huge memory requirements since the data point should be stored in a very long vector having thousands of bits. To alleviate this problem, we propose a novel HDC architecture, called StrideHD. By utilizing the window striding in image classification, StrideHD enables HDC system to be trained and tested using binary hypervectors and achieves high accuracy with fast training speed and significantly low hardware resources. StrideHD encodes data points to distributed binary hypervectors and eliminates the expensive Channel item Memory (CiM) and item Memory (iM) in the encoder, which significantly reduces the required hardware cost for inference. Our evaluation also shows that compared with two popular HD algorithms, the singlepass StrideHD model achieves a 27.6<inline-formula> <tex-math notation="LaTeX">$\times$ </tex-math></inline-formula> and 8.2<inline-formula> <tex-math notation="LaTeX">$\times$ </tex-math></inline-formula> reduction in inference memory cost without hurting the classification accuracy, while the iterative mode further provides 8.7<inline-formula> <tex-math notation="LaTeX">$\times$ </tex-math></inline-formula> memory efficiency. Under the same inference memory cost, our single-pass mode StrideHD averagely achieves 13.56% accuracy improvement in comparison with the single-pass baseline HD, which is a similar performance even in comparison with the costly iterative baseline HD models. As an extension, the iterative retraining mode of StrideHD averagely provides 11.33% accuracy improvement to its single-pass mode, which can be accomplished in fewer iterations in comparison with the baseline HD algorithms. The hardware implementation also demonstrates that StrideHD achieves over 9.9<inline-formula> <tex-math notation="LaTeX">$\times$ </tex-math></inline-formula> and 28.8<inline-formula> <tex-math notation="LaTeX">$\times$ </tex-math></inline-formula> reduction compared with baseline in area and power, respectively. |
| format | Article |
| id | doaj-art-91f45b301e794ddc964ba2a1d0a40a62 |
| institution | Kabale University |
| issn | 2644-1225 |
| language | English |
| publishDate | 2024-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Open Journal of Circuits and Systems |
| spelling | doaj-art-91f45b301e794ddc964ba2a1d0a40a622025-01-21T00:02:47ZengIEEEIEEE Open Journal of Circuits and Systems2644-12252024-01-01521122310.1109/OJCAS.2024.340102810530353StrideHD: A Binary Hyperdimensional Computing System Utilizing Window Striding for Image ClassificationDehua Liang0https://orcid.org/0000-0002-4922-3921Jun Shiomi1https://orcid.org/0000-0003-2733-9349Noriyuki Miura2https://orcid.org/0000-0002-0072-6114Hiromitsu Awano3https://orcid.org/0000-0002-3674-4584Graduate School of Information Science and Technology, Osaka University, Suita, Osaka, JapanGraduate School of Information Science and Technology, Osaka University, Suita, Osaka, JapanGraduate School of Information Science and Technology, Osaka University, Suita, Osaka, JapanGraduate School of Informatics, Kyoto University, Kyoto, JapanHyper-Dimensional (HD) computing is a brain-inspired learning approach for efficient and fast learning on today’s embedded devices. HDC first encodes all data points to high-dimensional vectors called hypervectors and then efficiently performs the classification task using a well-defined set of operations. Although HDC achieved reasonable performances in several practical tasks, it comes with huge memory requirements since the data point should be stored in a very long vector having thousands of bits. To alleviate this problem, we propose a novel HDC architecture, called StrideHD. By utilizing the window striding in image classification, StrideHD enables HDC system to be trained and tested using binary hypervectors and achieves high accuracy with fast training speed and significantly low hardware resources. StrideHD encodes data points to distributed binary hypervectors and eliminates the expensive Channel item Memory (CiM) and item Memory (iM) in the encoder, which significantly reduces the required hardware cost for inference. Our evaluation also shows that compared with two popular HD algorithms, the singlepass StrideHD model achieves a 27.6<inline-formula> <tex-math notation="LaTeX">$\times$ </tex-math></inline-formula> and 8.2<inline-formula> <tex-math notation="LaTeX">$\times$ </tex-math></inline-formula> reduction in inference memory cost without hurting the classification accuracy, while the iterative mode further provides 8.7<inline-formula> <tex-math notation="LaTeX">$\times$ </tex-math></inline-formula> memory efficiency. Under the same inference memory cost, our single-pass mode StrideHD averagely achieves 13.56% accuracy improvement in comparison with the single-pass baseline HD, which is a similar performance even in comparison with the costly iterative baseline HD models. As an extension, the iterative retraining mode of StrideHD averagely provides 11.33% accuracy improvement to its single-pass mode, which can be accomplished in fewer iterations in comparison with the baseline HD algorithms. The hardware implementation also demonstrates that StrideHD achieves over 9.9<inline-formula> <tex-math notation="LaTeX">$\times$ </tex-math></inline-formula> and 28.8<inline-formula> <tex-math notation="LaTeX">$\times$ </tex-math></inline-formula> reduction compared with baseline in area and power, respectively.https://ieeexplore.ieee.org/document/10530353/Hyperdimensional computingdistributed systemmemory requirement |
| spellingShingle | Dehua Liang Jun Shiomi Noriyuki Miura Hiromitsu Awano StrideHD: A Binary Hyperdimensional Computing System Utilizing Window Striding for Image Classification IEEE Open Journal of Circuits and Systems Hyperdimensional computing distributed system memory requirement |
| title | StrideHD: A Binary Hyperdimensional Computing System Utilizing Window Striding for Image Classification |
| title_full | StrideHD: A Binary Hyperdimensional Computing System Utilizing Window Striding for Image Classification |
| title_fullStr | StrideHD: A Binary Hyperdimensional Computing System Utilizing Window Striding for Image Classification |
| title_full_unstemmed | StrideHD: A Binary Hyperdimensional Computing System Utilizing Window Striding for Image Classification |
| title_short | StrideHD: A Binary Hyperdimensional Computing System Utilizing Window Striding for Image Classification |
| title_sort | stridehd a binary hyperdimensional computing system utilizing window striding for image classification |
| topic | Hyperdimensional computing distributed system memory requirement |
| url | https://ieeexplore.ieee.org/document/10530353/ |
| work_keys_str_mv | AT dehualiang stridehdabinaryhyperdimensionalcomputingsystemutilizingwindowstridingforimageclassification AT junshiomi stridehdabinaryhyperdimensionalcomputingsystemutilizingwindowstridingforimageclassification AT noriyukimiura stridehdabinaryhyperdimensionalcomputingsystemutilizingwindowstridingforimageclassification AT hiromitsuawano stridehdabinaryhyperdimensionalcomputingsystemutilizingwindowstridingforimageclassification |