Journal article · 2026 · Open access
EMERY: A Distribution-Matched Quantization and Near-Memory Heterogeneous Accelerator on FPGA for Sparse CNN Inference
IEEE Transactions on Neural Networks and Learning Systems, PP, 1–15 · Published 12 November 2026 · DOI 10.1109/tnnls.2026.3749815
In brief
“EMERY: A Distribution-Matched Quantization and Near-Memory Heterogeneous Accelerator on FPGA for Sparse CNN Inference” is a 2026 journal article by Bowen Yao and 4 co-authors, published in IEEE Transactions on Neural Networks and Learning Systems. OpenAlex counts 0 citations as of 18 November 2026. It is open access under a CC BY licence: the full text is free to read at IEEE Transactions on Neural Networks and Learning Systems.
- Citations
- 0
- Authors
- 5
Abstract
In resource-constrained edge devices, the pruning and quantization techniques are commonly employed together to minimize the computational and memory requirements of convolutional neural networks (CNNs). However, quantizing sparse CNNs still poses significant challenges, particularly regarding accuracy loss and computational performance optimization. These challenges primarily arise from the nonuniform distribution of sparse data, the limitations of existing quantization strategies, and the mismatch between hardware resources and the sparse weights of CNNs. To address these challenges, this article proposes a mixed-strategy quantization method for sparse distribution matching and its acceleration technique. By establishing a learning mechanism for sparse CNNs under joint constraints of low-bit-width parameters and heterogeneous field-programmable gate array (FPGA) resources, this method enhances the expressive power of sparse CNNs in low-bit-width data domains and improves computational efficiency on heterogeneous FPGA resources. First, by leveraging the learning mechanism of CNNs, low-bit-width quantization is automatically adapted to the irregular characteristics of sparse weights, thereby reducing quantization errors with traditional uniform strategies. Second, an FPGA-based heterogeneous near-memory computing architecture is designed to align with the computational characteristics of the quantized model. This architecture enhances compatibility between heterogeneous resources and algorithms through a mixed shift-and-multiply operation, while fully leveraging the FPGA's resource advantages. Furthermore, an allocation strategy that co-optimizes heterogeneous hardware and the quantization method is proposed to improve the overall energy efficiency. Finally, compared with the related works, the results show that the proposed quantization method reduces the accuracy loss by 30.43%-95.03% and improves the compression ratio by $3.57\times \sim 127.88\times $ on VGG-16, ResNet-18, and MobileNet-v2. In addition, the proposed accelerator achieves improvements in throughput and energy efficiency by an average of $4.56\times $ and $4.69\times $ , respectively.
Cite this paper
Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.
APA 7
Yao, B., Peng, X., Liu, L., Zhang, H., & Peng, Y. (2026). EMERY: A Distribution-Matched Quantization and Near-Memory Heterogeneous Accelerator on FPGA for Sparse CNN Inference. IEEE Transactions on Neural Networks and Learning Systems, PP, 1–15. https://doi.org/10.1109/tnnls.2026.3749815
MLA 9
Yao, Bowen, et al. “EMERY: A Distribution-Matched Quantization and Near-Memory Heterogeneous Accelerator on FPGA for Sparse CNN Inference.” IEEE Transactions on Neural Networks and Learning Systems, vol. PP, 2026, pp. 1–15. https://doi.org/10.1109/tnnls.2026.3749815.
Chicago
Yao, Bowen, Xiyuan Peng, Liansheng Liu, Haijun Zhang, and Yu Peng. 2026. “EMERY: A Distribution-Matched Quantization and Near-Memory Heterogeneous Accelerator on FPGA for Sparse CNN Inference.” IEEE Transactions on Neural Networks and Learning Systems PP: 1–15. https://doi.org/10.1109/tnnls.2026.3749815.
Harvard
Yao, B. et al. (2026) ‘EMERY: A Distribution-Matched Quantization and Near-Memory Heterogeneous Accelerator on FPGA for Sparse CNN Inference’, IEEE Transactions on Neural Networks and Learning Systems, PP, pp. 1–15. doi:10.1109/tnnls.2026.3749815.
IEEE
B. Yao, X. Peng, L. Liu, H. Zhang, and Y. Peng, “EMERY: A Distribution-Matched Quantization and Near-Memory Heterogeneous Accelerator on FPGA for Sparse CNN Inference,” IEEE Transactions on Neural Networks and Learning Systems, vol. PP, pp. 1–15, 2026, doi: 10.1109/tnnls.2026.3749815.
BibTeX
@article{yao2026emery,
title = {{EMERY: A Distribution-Matched Quantization and Near-Memory Heterogeneous Accelerator on FPGA for Sparse CNN Inference}},
author = {Yao, Bowen and Peng, Xiyuan and Liu, Liansheng and Zhang, Haijun and Peng, Yu},
journal = {IEEE Transactions on Neural Networks and Learning Systems},
year = {2026},
volume = {PP},
pages = {1--15},
doi = {10.1109/tnnls.2026.3749815},
url = {https://doi.org/10.1109/tnnls.2026.3749815}
}
Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)
Topic and keywords
Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- near-memory computing
- FPGA accelerator
- heterogeneous accelerators
- edge inference
- energy efficiency
- model compression
Work with this paper
Most-cited papers in Computer Science
- ImageNet classification with deep convolutional neural networksAlex Krizhevsky, Ilya Sutskever, Geoffrey E. Hinton · Communications of the ACM · 2017 · 114,726 citations
- Long Short-Term MemorySepp Hochreiter, Jürgen Schmidhuber · Neural Computation · 1997 · 109,097 citations
- A Mathematical Theory of CommunicationClaude E. Shannon · Bell System Technical Journal · 1948 · 91,736 citations
- Deep learningYann LeCun, Yoshua Bengio, Geoffrey E. Hinton · Nature · 2015 · 90,226 citations
- Faster R-CNN: Towards Real-Time Object Detection with Region Proposal NetworksShaoqing Ren, Kaiming He, Ross Girshick and 1 more · IEEE Transactions on Pattern Analysis and Machine Intelligence · 2016 · 74,143 citations
Questions about this paper
How do I cite this paper in APA style?
Yao, B., Peng, X., Liu, L., Zhang, H., & Peng, Y. (2026). EMERY: A Distribution-Matched Quantization and Near-Memory Heterogeneous Accelerator on FPGA for Sparse CNN Inference. IEEE Transactions on Neural Networks and Learning Systems, PP, 1–15. https://doi.org/10.1109/tnnls.2026.3749815
How many times has this paper been cited?
0 times, according to OpenAlex on 18 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.
Is this paper free to read?
Yes. It is open access at IEEE Transactions on Neural Networks and Learning Systems: sign in and open the full text from this page.
Who wrote this paper?
Bowen Yao, Xiyuan Peng, Liansheng Liu, Haijun Zhang, Yu Peng.
What is the DOI of this paper?
10.1109/tnnls.2026.3749815, permanently at https://doi.org/10.1109/tnnls.2026.3749815. Sign in and open it from this page.
Bibliographic data from OpenAlex (CC0), last checked 18 November 2026. ResearchGuru does not host this paper; once you are signed in, the links open it at the publisher or the repository that does.