Clo-HDnn: A 4.66 TFLOPS/W and 3.78 TOPS/W Continual On-Device Learning Accelerator with Energy-efficient Hyperdimensional Computing via Progressive Search
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916861069754368 |
|---|---|
| author | Song, Chang Eun Xu, Weihong Fan, Keming Jain, Soumil Hota, Gopabandhu Yang, Haichao Liu, Leo Akarvardar, Kerem Chang, Meng-Fan Diaz, Carlos H. Cauwenberghs, Gert Rosing, Tajana Kang, Mingu |
| author_facet | Song, Chang Eun Xu, Weihong Fan, Keming Jain, Soumil Hota, Gopabandhu Yang, Haichao Liu, Leo Akarvardar, Kerem Chang, Meng-Fan Diaz, Carlos H. Cauwenberghs, Gert Rosing, Tajana Kang, Mingu |
| contents | Clo-HDnn is an on-device learning (ODL) accelerator designed for emerging continual learning (CL) tasks. Clo-HDnn integrates hyperdimensional computing (HDC) along with low-cost Kronecker HD Encoder and weight clustering feature extraction (WCFE) to optimize accuracy and efficiency. Clo-HDnn adopts gradient-free CL to efficiently update and store the learned knowledge in the form of class hypervectors. Its dual-mode operation enables bypassing costly feature extraction for simpler datasets, while progressive search reduces complexity by up to 61% by encoding and comparing only partial query hypervectors. Achieving 4.66 TFLOPS/W (FE) and 3.78 TOPS/W (classifier), Clo-HDnn delivers 7.77x and 4.85x higher energy efficiency compared to SOTA ODL accelerators. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2507_17953 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Clo-HDnn: A 4.66 TFLOPS/W and 3.78 TOPS/W Continual On-Device Learning Accelerator with Energy-efficient Hyperdimensional Computing via Progressive Search Song, Chang Eun Xu, Weihong Fan, Keming Jain, Soumil Hota, Gopabandhu Yang, Haichao Liu, Leo Akarvardar, Kerem Chang, Meng-Fan Diaz, Carlos H. Cauwenberghs, Gert Rosing, Tajana Kang, Mingu Hardware Architecture Machine Learning Clo-HDnn is an on-device learning (ODL) accelerator designed for emerging continual learning (CL) tasks. Clo-HDnn integrates hyperdimensional computing (HDC) along with low-cost Kronecker HD Encoder and weight clustering feature extraction (WCFE) to optimize accuracy and efficiency. Clo-HDnn adopts gradient-free CL to efficiently update and store the learned knowledge in the form of class hypervectors. Its dual-mode operation enables bypassing costly feature extraction for simpler datasets, while progressive search reduces complexity by up to 61% by encoding and comparing only partial query hypervectors. Achieving 4.66 TFLOPS/W (FE) and 3.78 TOPS/W (classifier), Clo-HDnn delivers 7.77x and 4.85x higher energy efficiency compared to SOTA ODL accelerators. |
| title | Clo-HDnn: A 4.66 TFLOPS/W and 3.78 TOPS/W Continual On-Device Learning Accelerator with Energy-efficient Hyperdimensional Computing via Progressive Search |
| topic | Hardware Architecture Machine Learning |
| url | https://arxiv.org/abs/2507.17953 |