Fused-layer CNN accelerators
The DOI matches its Crossref record; 20 of the 20 references it cites were found in a registry or an index.
Abstract
Deep convolutional neural networks (CNNs) are rapidly becoming the dominant approach to computer vision and a major component of many other pervasive machine learning tasks, such as speech recognition, natural language processing, and fraud detection. As a result, accelerators for efficiently evaluating CNNs are rapidly growing in popularity. The conventional approaches to designing such CNN accelerators is to focus on creating accelerators to iteratively process the CNN layers. However, by processing each layer to completion, the accelerator designs must use off-chip memory to store intermediate data between layers, because the intermediate data are too large to fit on chip. In this work, we observe that a previously unexplored dimension exists in the design space of CNN accelerators that focuses on the dataflow across convolutional layers. We find that we are able to fuse the processing of multiple CNN layers by modifying the order in which the input data are brought on chip, enabling caching of intermediate data between the evaluation of adjacent CNN layers. We demonstrate the effectiveness of our approach by constructing a fused-layer CNN accelerator for the first five convolutional layers of the VGGNet-E network and comparing it to the state-of-the-art accelerator implemented on a Xilinx Virtex-7 FPGA. We find that, by using 362KB of on-chip storage, our fused-layer accelerator minimizes off-chip feature map data transfer, reducing the total transfer by 95%, from 77MB down to 3.6MB per image.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“When the number of worker nodes increases from 2 to 4, MoDNN can speedup the DNN computation by 2.17-4.28×[130]”
From Convergence of Edge Computing and Deep Learning: A Comprehensive Survey (2020) · cites this paper as [130]
In the citing paper
“With regard to specific DNN structures, e.g., CNN, a finer grid partitioning can be applied to minimize communication, synchronization, and memory overhead [130].”
From Convergence of Edge Computing and Deep Learning: A Comprehensive Survey (2020) · cites this paper as [130]
Citations
Cites
Very Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148 · cited 6×
“Table II shows the results of a similar comparison, but for the larger and newer VGGNet-E network [17].”
BackgroundMethodResultInfluentialImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155 · cited 4×
“(We observe a similar pattern in other common CNN structures, such as AlexNet [8].”
BackgroundMethodInfluentialGoing deeper with convolutions
Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., & Rabinovich, A. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298594 · Cited by 48,562
ImageNet Large Scale Visual Recognition Challenge
Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al. (2015) · International Journal of Computer Vision · doi:10.1007/s11263-015-0816-y · Cited by 39,380 · cited 3×
“Large Scale Visual Recognition Challenge (ILSVRC) [15], uses five convolutional layers, while VGG [17] (winner of a portion of ILSVRC 2014) uses up to 17 convolutional layers.”
MethodOptimizing FPGA-based Accelerator Design for Deep Convolutional Neural Networks
Zhang, C., Li, P., Sun, G., Guan, Y., Xiao, B., & Cong, J. (2015) · Proceedings of the 2015 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays · doi:10.1145/2684746.2689060 · Cited by 2,158 · cited 10×
“The computational dominance of the convolutional layers, coupled with the need for larger and deeper networks, has sparked significant interest in the design and optimization of accelerator structures for these layers [19], [10].”
BackgroundMethodResultInfluentialDaDianNao: A Machine-Learning Supercomputer
Chen, Y., Luo, T., Liu, S., Zhang, S., He, L., Wang, J., Li, L., Chen, T., Xu, Z., Sun, N., et al. (2014) · 2014 47th Annual IEEE/ACM International Symposium on Microarchitecture · doi:10.1109/micro.2014.58 · Cited by 1,492 · cited 3×
“As a result, there has been significant interest in developing and adapting hardware accelerators for CNNs [7] such as GPUs [3], FPGAs [13], [19], [10], [1], and ASICs [2].”
BackgroundMethodTorch7: A Matlab-like Environment for Machine Learning
Collobert, R., Kavukcuoglu, K., & Farabet, C. (2011) · Infoscience (Ecole Polytechnique Fédérale de Lausanne) · Cited by 1,186
cuDNN: Efficient Primitives for Deep Learning
Chetlur, S., Woolley, C., Vandermersch, P., Cohen, J., Tran, J. J., Catanzaro, B., & Shelhamer, E. (2014) · arXiv (Cornell University) · doi:10.48550/arxiv.1410.0759 · Cited by 943 · cited 2×
“As a result, there has been significant interest in developing and adapting hardware accelerators for CNNs [7] such as GPUs [3], FPGAs [13], [19], [10], [1], and ASICs [2].”
MethodA reconfigurable fabric for accelerating large-scale datacenter services
Putnam, A., Caulfield, A. M., Chung, E. S., Chiou, D., Constantinides, K., Demme, J., Esmaeilzadeh, H., Fowers, J., Gopal, G. P., Gray, J., et al. (2014) · 2014 ACM/IEEE 41st International Symposium on Computer Architecture (ISCA) · doi:10.1109/isca.2014.6853195 · Cited by 576
NeuFlow: A runtime reconfigurable dataflow processor for vision
Farabet, C., Martini, B., Corda, B., Akselrod, P., Culurciello, E., & LeCun, Y. (2011) · CVPR 2011 WORKSHOPS · doi:10.1109/cvprw.2011.5981829 · Cited by 369 · cited 1×
“In [6], [5], [16], systolic implementations were proposed.”
Cited by
Items 601–700 of 779
CADOSys: Cache Aware Design Space Optimization for Spatial ML Accelerators
Li, R., Ma, S., Kavi, K. M., Mehta, G., Yadwadkar, N. J., & John, L. K. (2025) · ACM Great Lakes Symposium on VLSI (GLSVLSI) · doi:10.1145/3716368.3735222 · Cited by 1
Intra- and Inter-Layer Scheduling Exploration and Optimization for ReRAM-Based DNN Accelerators
Zhao, Y., Ma, S., Li, T., Zhang, J., & Tang, Y. (2025) · IEEE Transactions on Computers · doi:10.1109/tc.2025.3584270 · Cited by 1
“For example, the fused-layer CNN accelerator [16] introduces a pyramid-shaped multi-layer sliding window for realizing the input activation reuse for different layers.”
PIMapping: A Tile-Level Dataflow Optimization Framework for PIM Architecture
Zhu, Z., Zhou, Y., Zhu, J., Wang, Y., & Pan, H. (2025) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2025.3621513 · Cited by 1
“In a data-driven execution model [24], [39], this can lead to potential contention for interconnect resources.”
IsoSched: Preemptive Tile Cascaded Scheduling of Multi-DNN via Subgraph Isomorphism
Zhao, B., Yuan, Z., Liu, H., Hu, Y., Zhai, H., Zhang, H., Zhao, W., Xia, T., & Ren, P. (2026) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2026.3709688 · Cited by 1
MCExplorer: Exploring the Design Space of Multiple Compute-Engine Deep Learning Accelerators
Qararyah, F. M., Maleki, M. A., & Trancoso, P. (2025) · ACM Transactions on Architecture and Code Optimization · doi:10.1145/3774913 · Cited by 1
“These results show that MC E xplorer identified accelerators that better utilize the PEs, resulting in higher performance densities.”
Hierarchical-ISA Supporting Row-Wise Operands for Efficient DNN Computation
Huo, Z., Zhao, W., Dang, Q., Ma, C., Yang, G., Fu, G., Xia, T., & Ren, P. (2025) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2025.3637162 · Cited by 1
“To address the challenges of indexing complexity and fragmentation, Alwani et al. [16] copied the reusable data from the shared buffer to a dedicated buffer.”
InfluentialAn Energy-Efficient Super-Resolution Processor with Enhanced Tiling Artifact Reduction
Yoo, B., Park, S., Lee, S., & Jeon, D. (2025) · IEEE Asia Pacific Conference on Circuits and Systems (APCCAS) · doi:10.1109/apccas67402.2025.11377884 · Cited by 1
“1(b) addresses this challenge by splitting the input into smaller blocks and temporarily store each block in on-chip memory, effectively minimizing off-chip memory access [4].”
Nie, Z., Wang, H., Chronopoulos, A. T., Tang, Z., Li, K., Liu, C., & Xiao, Z. (2026) · IEEE Transactions on Parallel and Distributed Systems · doi:10.1109/tpds.2026.3667207 · Cited by 1
“Prior works [5], [6], [7], [8], [9], [10], [11] have demonstrated that block-wise mapping is a promising strategy for this challenge.”
Detecting Phishing Attacks Using Recurrent Neural Networks and a Hybrid LSTM-RNN Model
Sabeeh, F., & Alazzawi, A. (2026) · Lecture notes in networks and systems · doi:10.1007/978-3-032-27669-8_4 · Cited by 1
Optimization through recomputation in the polyhedral model
Jongen, M. H., Waeijen, L., Jordans, R., Jóźwiak, L., & Corporaal, H. (2018) · TU/e Research Portal · Cited by 1
Liu, C., Yang, H., Wang, R., Luan, Z., & Qian, D. (2018) · Future Generation Computer Systems · doi:10.1016/j.future.2018.02.024 · Cited by 1
Heuristic Approximation of Early-Stage CNN Data Representation for Vision Intelligence Systems
Choi, J., Sampson, J., & Narayanan, V. (2018) · IEEE International Conference on Computer Design: VLSI in Computers and Processors (ICCD) · doi:10.1109/iccd.2018.00041 · Cited by 1
“Additional cache can reduce off-chip memory traffic by fusing CNN layers in intermediate data locality [10].”
BackgroundElasto-Net: An HDL Conversion Framework For Convolutional Neural Networks
Ansari, A., & Ogunfunmi, T. (2018) · Asilomar Conference on Signals, Systems and Computers (ACSSC) · doi:10.1109/acssc.2018.8645344 · Cited by 1
Performance improvement methods
Takano, S. (2021) · Elsevier eBooks · doi:10.1016/b978-0-12-818279-6.00016-5 · Cited by 1
Nazir, A., Mir, R. N., & Qureshi, S. (2021) · Concurrency and Computation Practice and Experience · doi:10.1002/cpe.6593 · Cited by 1
Intra- and Inter- Layer Transformation to Reduce Memory Traffic for CNN Computation
Liao, P.-W., Hsu, W., & Liao, S.-W. (2021) · International Conference on Parallel Processing Workshops (ICPPW) · doi:10.1145/3458744.3473353 · Cited by 1
“[1] proposes fused-layer CNN Accelerators.”
Reconfigurable Architecture and Dataflow for Memory Traffic Minimization of CNNs Computation
Cheng, W., Liu, X., Wu, H.-T., Pai, H.-Y., & Chung, P.-Y. (2021) · Micromachines · doi:10.3390/mi12111365 · Cited by 1
“Research [2,3] proposed dataflow enhancement techniques.”
MethodFusion for Tile-based Deconvolution Layers
Jeong, M.-W., & Rhee, C. E. (2021) · International SoC Design Conference · doi:10.1109/isocc53507.2021.9613947 · Cited by 1
“There have been many studies to improve energy efficiency by minimizing intermediate data movement between on-chip and off-chip by utilizing the characteristic of the encoder part where the size of the activation map between layers gradually decreases due to pooling [5][6].”
BackgroundPark, P., Lee, J., Jeong, H., & Kim, J. (2022) · IEEE Transactions on Parallel and Distributed Systems · doi:10.1109/tpds.2022.3144453 · Cited by 1
“Note that DLS offloads only unsupported or in-efficient operations, so that optimizations in ASICs (e.g., layer fusion [26]) for supported Mat and Aux operations are still applicable.”
BackgroundYang, R., Shen, J., Wen, M., Cao, Y., & Li, Y. (2022) · Electronics · doi:10.3390/electronics11101587 · Cited by 1
Zhao, Y., Chen, Y., & Xu, Z. (2022) · Intelligent Computing · doi:10.34133/2022/9797623 · Cited by 1
“Moreover, the developer may also need to implement fused operators [3] for optimal efficiency, contributing to a significant part of the development workload.”
BackgroundA Fine-Grained Modeling Approach for Systolic Array-Based Accelerator
Li, Y., Wen, M., Fei, J., Shen, J., & Cao, Y. (2022) · Electronics · doi:10.3390/electronics11182928 · Cited by 1
DefT: Boosting Scalability of Deformable Convolution Operations on GPUs
Hanson, E., Horton, M., Li, H. H., & Chen, Y. (2023) · International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS) · doi:10.1145/3582016.3582017 · Cited by 1
“CNN inter-layer fusion [3] exploits the sequential nature of CNN layers by immediately consuming IFM elements as they are produced by a previous layer.”
BackgroundConvolutional Neural Network implemented on FPGA for trajectory classification
Feldioreanu, G., Popa, S., & Ivanovici, M. (2023) · International Symposium on Signals, Circuits and Systems · doi:10.1109/isscs58449.2023.10190948 · Cited by 1
“In [4] a method to optimize memory access overhead by fusing CNN layers, targeting a Xilinx Virtex FPGA is proposed.”
MethodACCO: Automated Causal CNN Scheduling Optimizer for Real-Time Edge Accelerators
Yin, J., Mei, L., Guntoro, A., & Verhelst, M. (2023) · IEEE International Conference on Computer Design: VLSI in Computers and Processors (ICCD) · doi:10.1109/iccd58817.2023.00065 · Cited by 1
“Spatial CNNs can exploit DF execution (a.k.a layer fusion [8]).”
BackgroundMethodDadras, I., Seydi, S., Ahmadilivani, M. H., Raik, J., & Salehi, M. E. (2023) · IEEE International Conference on Electronics, Circuits and Systems (ICECS) · doi:10.1109/icecs58634.2023.10382831 · Cited by 1
“There are endeavors to fit the networks on the devices by reducing bit-width [2], fusing subsequent layers [3], and utilizing approximate computing [4], in-memory computing [5], and analog computing [6].”
BackgroundEnabling Ultra-Low Power Ultrasound Imaging with Compute-in-Memory Sparse Reconstruction Accelerator
Li, W., Zhang, X., Lee, J., Degertekin, F. L., Li, S., & Yu, S. (2023) · IEEE Biomedical Circuits and Systems Conference (BioCAS) · doi:10.1109/biocas58349.2023.10389000 · Cited by 1
“The resulting input halo phenomenon limits the depth of inter-layer pipelining [31] and is especially problematic for processing low-resolution input images such as used in portable ultrasound.”
BackgroundA 3D Hybrid Optical-Electrical NoC Using Novel Mapping Strategy Based DCNN Dataflow Acceleration
Zhang, B., Gu, H., Zhang, G. L., Yang, Y., Ma, Z., & Schlichtmann, U. (2024) · IEEE Transactions on Parallel and Distributed Systems · doi:10.1109/tpds.2024.3394747 · Cited by 1
“At present, some DCNN accelerators based on dataflow have been proposed to reduce the number of data movement by utilizing the reusability of data, thus realizing DCNN acceleration and reducing EC [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33].”
MethodTENG: A General-Purpose and Efficient Processor Architecture for Accelerating DNN
Zhang, Z., Cai, Y., Liao, T., Xu, C., & Jiao, X. (2024) · IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS) · doi:10.1109/aicas59952.2024.10595854 · Cited by 1
“+ Element-Wise, also known as operator fusion [15], [16], we map operator fusion to ME and VE to complete the operation, as shown in Fig.”
MethodHigh-Rate. Compact In-Sensor Denoising for Active Stereo Vision Towards Embedded Depth Sensing
Houshmand, P., Staelens, J.-S., Tempel, W. V. D., & Verhelst, M. (2024) · IEEE Interregional NEWCAS Conference (NEWCAS) · doi:10.1109/newcas58973.2024.10666331 · Cited by 1
“After the synchronized projection layer an array of processing elements (PEs) executes the convolution stage on the incoming stream of projections, following a line-buffered, depth-first dataflow [23], [24], as in Fig.”
MethodDeep Reorganization: Retaining Residuals in TinyML
Mendis, H. R., Kang, C.-K., Lin, C., Chen⋆, M.-S., & Hsiu, P.-C. (2024) · ACM/IEEE Design Automation Conference (DAC) · doi:10.1145/3649329.3656219 · Cited by 1
“For YOLOv5n, as shown in Figure 2(b), peak memory usage can be reduced by employing depth-first inference [1, 13], where the output feature map of a layer is divided into smaller patches and convolution is performed on the receptive field of each patch, rather than computing layer-by-layer.”
PODALA: Power-Efficient Object Detection Accelerator With Customized Layer Fusion Engine
Yue, T., Chang, L., Xu, H., Wang, C., Lin, S., & Zhou, J. (2024) · Integrated Circuits and Systems · doi:10.23919/ics.2024.3506511 · Cited by 1
Li, M., Lu, S., & Han, J. S. (2024) · IEEE International Conference on Solid-State and Integrated Circuit Technology · doi:10.1109/icsict62049.2024.10831247 · Cited by 1
A Deep Learning Accelerator for Modified YOLOv7-tiny with Shortcut-Aware Layer Fusion
Huang, W.-E., Huang, Y.-S., & Jou, S.-J. J. (2025) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas56072.2025.11043183 · Cited by 1
An Analytical Cost Model for Fast Evaluation of Multiple Compute-Engine CNN Accelerators
Qararyah, F. M., Maleki, M. A., & Trancoso, P. (2025) · IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS) · doi:10.1109/ispass64960.2025.00030 · Cited by 1
“A systematic accelerator design approach requires analyzing the bottlenecks and inefficiencies and exploring the design space to identify optimizations that mitigate them.”
Lu, T.-C., Ding, J., Jiang, B., Zhang, H., Xu, W., & Chai, Z. (2025) · ACM Transactions on Reconfigurable Technology and Systems · doi:10.1145/3770759 · Cited by 1
“This progress comes with severe deployment challenges for edge devices: complex DNN operators like high-dimensional convolutions dominate hardware resources [1].”
Distributed Convolutional Neural Network Training on Mobile and Edge Clusters
Rama, P., Threadgill, M., & Gerstlauer, A. (2025) · IFIP advances in information and communication technology · doi:10.1007/978-3-032-07102-6_6 · Cited by 1
Zhou, L. (2019) · OhioLink ETD Center (Ohio Library and Information Network) · Cited by 1
Efficient Implementations of Sparse and Quantized Deep Neural Networks Using Systolic Arrays
McDanel, B. (2019) · Digital Access to Scholarship at Harvard (DASH) (Harvard University) · Cited by 1
Loom and Laconic: Hardware Accelerators for Deep Learning Algorithms
Sharifymoghaddam, S. (2020) · TSpace (University of Toronto) · Cited by 1
Scaling Up Deep Neural Network Optimization for Edge InferencePreprint
Lu, B., Yang, J., & Ren, S. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2009.00278 · Cited by 1
“There have also been many prior studies on DNN model compression, such as pruning and quantization [1,10,11,14,17,18,22,23,27,30], matrix factorization [13,28], and knowledge distillation [32], among others.”
Architecture and System Support for Safety-aware Autonomous Vehicle Design
Zhao, H. (2021) · eScholarship (California Digital Library) · Cited by 1
SIMCNN - Exploiting Computational Similarity to Accelerate CNN Training in Hardware.Preprint
Janfaza, V., Weston, K., Razavi, M., Mandal, S., & Muzahid, A. A. M. (2021) · arXiv (Cornell University) · Cited by 1
Phantom: A High-Performance Computational Core for Sparse Convolutional Neural NetworksPreprint
Qureshi, M. A., & Munir, A. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2111.05002 · Cited by 1
“Efficient data reuse-based accelerators [2, 8] maximize the data reuse within different layers to minimize the memory accesses, thereby, reducing energy consumption.”
BackgroundHeterogeneous Multi-core Array-based DNN AcceleratorPreprint
Maleki, M. A., Kamal, M., & Afzali‐Kusha, A. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2206.12605 · Cited by 1
“Encoding and compression [25]–[27], using approximate computing [28], [29], utilizing quantized or truncated values in the processing of the networks [30]–[33], changing dataflow and re-ordering the computations to reduce the memory access can be classified in this category of works [22], [34].”
MethodA Formalism of DNN Accelerator FlexibilityPreprint
Kao, S.-C., Kwon, H., Pellauer, M., Parashar, A., & Krishna, T. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2206.02987 · Cited by 1
Vu, M. T., Garpebring, A., Nyholm, T., & Löfstedt, T. (2023) · doi:10.36227/techrxiv.21915921 · Cited by 1
A Fresh Perspective on DNN Accelerators by Performing Holistic Analysis Across ParadigmsPreprint
Glint, T., Jha, C. K., Awasthi, M., & Mekie, J. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2208.05294 · Cited by 1
“While there are a number of CHAs [23, 11, 16, 24, 25, 26, 27, 28], in this work, we have used the state-of-the-art, hardware realized Simba accelerator as a representative architecture for CHAs.”
BackgroundMethodInfluentialSF-MMCN: Low-Power Sever Flow Multi-Mode Diffusion Model AcceleratorPreprint
Hsu, H.-K., Wey, I., & Teo, T. H. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2403.10542 · Cited by 1
“Mo et al. [2021] proposed a dataflow inspired by Alwani et al. [2016] nad Mo et al. [2019].”
Taming TinyML: deep learning inference at computational extremes
Liberis, E. (2022) · Apollo (University of Cambridge) · doi:10.17863/cam.96947 · Cited by 1
A Real-time and Hardware Efficient Artefact-free Spike Sorting Using Deep Spike DetectionPreprint
Jiang, X., Fang, T., & Zamani, M. (2025) · doi:10.36227/techrxiv.174552724.43504042/v1 · Cited by 1
High performance deep learning on resource constrained platforms
Fernández-Marqués, J. (2021) · Oxford University Research Archive (ORA) (University of Oxford) · doi:10.5287/ora-7ryjdeebb · Cited by 1
Hardware Accelerators for Machine Learning Applications. Case Study : Autonomous Vehicles
Neggaz, M. A. (2020) · theses.fr (ABES) · doi:10.70675/09cf40c0z17a0z4efbz8c0fzb956e149d59c · Cited by 1
Nguyen, V. T. (2022) · theses.fr (ABES) · doi:10.70675/67883d2dz362cz4ca1z88f2z6a392d349bbb · Cited by 1
Efficient CNN inference acceleration on FPGAs : a pattern pruning-driven approach
Pradels, L. (2024) · theses.fr (ABES) · doi:10.70675/9b86665cz79bbz400fz9d0bz69433c967981 · Cited by 1
AccelForge: Comprehensive Modeling and Co-Design Framework for AI AcceleratorsPreprint
Andrulis, T., Gilbert, M., Sze, V., & Emer, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.11906 · Cited by 1
Valpreda, E. (2024) · PORTO@Iris · Cited by 1
An Optimized Architecture For Decomposed Convolutional Neural Networks
Sun, F., Lin, J. T., & Wang, Z. (2018) · IEEE Computer Society Annual Symposium on VLSI · doi:10.1109/isvlsi.2018.00100
Hardware-Accelerated Modeling of Large-Scale Networks-on-Chip
Chu, V. T., & Van, T. C. (2018)
Nie, Z., Li, Z., Wang, L., Guo, S., & Dou, Q. (2018) · Communications in computer and information science · doi:10.1007/978-981-13-2423-9_2
Eadnet: Efficient Architecture for Decomposed Convolutional Neural Networks
Sun, F., Lin, J. T., & Wang, Z. (2018) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp.2018.8461585
Bandwidth Efficient Architectures for Convolutional Neural Network
Wang, J., Lin, J. T., & Wang, Z. (2018) · SiPS ... design and implementation - IEEE Workshop on Signal Processing Systems · doi:10.1109/sips.2018.8598291
“[26] proposed an efficient dataflow across convolutional layers, which employed a pyramid-shaped multi-layer sliding window to enable effective on-chip caching during CNN evaluation.”
MethodOptimization Through Recomputation in the Polygedrak Model; Work in progress
Jongen, M., Waeijen, L., Jordans, R., Jóźwiak, L., & Corporaal, H. (2018) · Radboud Repository (Radboud University)
DeCoILFNet: Depth Concatenation and Inter-Layer Fusion based ConvNet AcceleratorPreprint
Baranwal, A., Bansal, I., Nahar, R., & Krishna, K. M. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1901.02774
“Traditional implementations of CNNs evaluate the network layer by layer [1] and off-load data intermittently to a larger external memory which significantly decreases throughput because of limited data transfer bandwidth.”
BackgroundMethodInfluentialLiu, H., Liu, L., Zhu, W., Li, Q., Mo, H., & Wei, S. (2019) · ACM/IEEE Design Automation Conference (DAC) · doi:10.1145/3316781.3317854
“2% than the baseline [9], achieving about 60% and 31% decrease in on-chip memory space and access respectively than the similar fused-layer accelerator [2].”
BackgroundMethodInfluentialImpact of CNNs Pooling Layer Implementation on FPGAs Accelerator Design
Muńío-Gracia, A., Fernández‐Berni, J., Carmona‐Galán, R., & Rodriguez-Vazquez, A. (2019) · International Conference on Distributed Smart Cameras · doi:10.1145/3349801.3357130
“Because of that, recent investigations about CNNs mapping into RTL focus on their execution optimization exploiting either inter-layer [5] or intra-layer [6] parallelization.”
BackgroundAttojoule scale computation of large optical neural networks
Sludds, A. (2019) · DSpace@MIT (Massachusetts Institute of Technology)
A Pattern-Based API for Mapping Applications to a Hierarchy of Multi-Core Devices
Guo, J., Teodorescu, R., & Agrawal, G. (2020) · IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (CCGrid) · doi:10.1109/ccgrid49817.2020.00-92
“In addition to load balancing multi-scale detection jobs, we also observe similar pyramid structure in fused-layer CNN [1], as is a potential generalization of this optimization.”
Ning, L. (2020) · NCSU Libraries Repository (North Carolina State University Libraries)
Artificial Intelligence Inference in Edge
Wang, X., Han, Y., Leung, V. C. M., Niyato, D. T., Yan, X., & Chen, X. (2020) · Springer eBooks · doi:10.1007/978-981-15-6186-3_5
Approximate Computing for Aging Compensation and Energy-efficient Neural Network
김희수 (2020) · Seoul National University Open Repository (Seoul National University)
A cross-stack, network-centric architectural design for next-generation datacenters
Alian, M. (2020) · Illinois Digital Environment for Access to Learning and Scholarship (University of Illinois at Urbana-Champaign)
Tile-based Code Generation for Efficiently Accessing to Scratchpad Memory
Lee, J., Paik, Y., & Kim, S. W. (2021) · International Conference on Electronics, Information, and Communication (ICEIC) · doi:10.1109/iceic51217.2021.9369731
Amplifying On-Chip Memory Storage Capacity via Compression for Neural Network Workloads
Edo, I. (2020) · TSpace (University of Toronto)
Online Training Refinement Network and Architecture Design for Stereo Matching
Wu, Y., Wu, S.-S., Huang, T., & Chen, L. (2021) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas51556.2021.9401468
Reducing Off-chip Memory Accesses in Deep Neural Network Accelerators
Siu, K. (2019) · TSpace (University of Toronto)
Roorda, E. (2021) · cIRcle (University of British Columbia) · doi:10.14288/1.0401092
Neural Network Instruction Set Extension and Code Mapping Mechanism
Lou, W., Wang, C., Gong, L., & Zhou, X. (2021) · International Journal of Software and Informatics · doi:10.21655/ijsi.1673-7288.00251
An Anatomization of FPGA-Based Neural Networks
Negi, A., Saxena, D., Kunal, & Suneja, K. (2021) · Lecture notes in networks and systems · doi:10.1007/978-981-16-2919-8_45
Globally Assisted Instance Normalization for Bandwidth-Efficient Neural Style Transfer
Hsu, H.-P., & Huang, C.-T. (2021) · SiPS ... design and implementation - IEEE Workshop on Signal Processing Systems · doi:10.1109/sips52927.2021.00019
“This kind of algorithm challenges state-of-the-art bandwidth-efficient CNN accelerators [7] [8].”
BackgroundAccelerating distributed neural network training with network-centric approach
Yuan, Y. (2019) · Illinois Digital Environment for Access to Learning and Scholarship (University of Illinois at Urbana-Champaign)
Optimization of Node-clustering-based DAG partition targeting NVDLA Architecture
Hu, S., Du, G., Li, J., Li, Z., Ni, W., Yongsheng, Y., & Wang, X. (2021) · IEEE International Conference on ASIC (ASICON) · doi:10.1109/asicon52560.2021.9620327
A Fused Inference Design for Pattern-Based Sparse CNN on Edge Devices
Guo, J., Teodorescu, R., & Agrawal, G. (2021) · IEEE International Conference on High Performance Computing, Data, and Analytics · doi:10.1109/hipc53243.2021.00060
(2021) · Elsevier eBooks · doi:10.1016/b978-0-12-818279-6.00025-6
(2021) · Elsevier eBooks · doi:10.1016/b978-0-12-818279-6.00005-0
Ding, C. (2019) · Digital Repository Service (Northeastern University) · doi:10.17760/d20323953
Memory and System Aware Architectures for Real-Time Machine Learning
Pinkham, R. (2021) · Deep Blue (University of Michigan) · doi:10.7302/3950
A Compressed Data Partition and Loop Scheduling Scheme for Neural Networks
Li, D., Fang, R., Wang, J., Zhao, D., Chong, T., Ren, Z., & Ma, J. (2022) · IEEE Access · doi:10.1109/access.2022.3204038
“Alwani et al. proposed caching the area shared by adja-175 cent receptive fields, so that only the nonshared part needs to be fetched from memory in the next layer calculation [21].”
BackgroundVu, M., Garpebring, A., Nyholm, T., & Löfstedt, T. (2023) · doi:10.36227/techrxiv.21915921.v1
Efficient, Reconfigurable, and QoS-Aware Systems for Deep Neural Networks
Wang, X. (2022) · Deep Blue (University of Michigan) · doi:10.7302/6872
Memory-Centric Fusion-based CNN Accelerator with 3D-NoC and 3D-DRAM
Lu, W., Ge, P.-Y., Huang, P.-T., Chen, H.-M., & Hwang, W. (2022) · International SoC Design Conference · doi:10.1109/isocc56007.2022.10031328
“The fusion-based 3D NoC is designed to realize efficient memory accesses and data communications among PEs by adopting a fused-layer technique [3] with stateof-the-art 3D stacked memory.”
MethodSelf-citationSynchronization-Aware NAS for an Efficient Collaborative Inference on Mobile Platforms
Kang, B. W., Wohn, J., Lee, S., Park, S., Noh, Y., & Park, Y. (2023) · ACM SIGPLAN/SIGBED Conference on Languages, Compilers, and Tools for Embedded Systems (LCTES) · doi:10.1145/3589610.3596284
“Sync Count: 3 Sync Position: [1,2,6] Sync Type: [‘s’,‘c’,‘s’] Sync Cost”
BackgroundDIANA: DIgital and ANAlog Heterogeneous Multi-core System-on-Chip
Jain, V., & Verhelst, M. (2023) · Springer eBooks · doi:10.1007/978-3-031-38230-7_7
Munir, A., Kong, J., & Qureshi, M. A. (2023) · Wiley eBooks · doi:10.1002/9781394171910.ref
Improvements in Interlayer Pipelining of CNN Accelerators Using Genetic AlgorithmsPreprint
Horeni, M., & Joshi, S. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2311.12235
“While recomputing is an option, previous works have found that caching is almost always better, so we employ that for future evaluations [11], [12], [14].”
BackgroundMethodInfluentialMonitor Placement for Fault Localization in Deep Neural Network AcceleratorsPreprint
Liu, W.-K. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2311.16594
“To address the intensive computational demands of DNNs, there is growing interest in specialized hardware accelerators for DNNs to enhance performance and energy efficiency [6]–[9].”
BackgroundOptTc: A method of optimizating memory access latency for convolutional acceleratorsPreprint
Zhao, H., Wang, Y., Fang, Z., & Wu, J. J. (2023) · doi:10.36227/techrxiv.170326747.73509974/v1
Miao, Y., & Ikeda, M. (2024) · IEICE Electronics Express · doi:10.1587/elex.21.20240206
Convolution tensor slicing optimization for multicore NPUs
Sousa, R. C. F. (2023) · doi:10.47749/t/unicamp.2023.1374643
Metadata
- Authors
- Alwani, M., Chen, H., Ferdman, M., & Milder, P.
- Affiliations
- Stony Brook University (US)
- Venue
- 2016 49th Annual IEEE/ACM International Symposium on Microarchitecture (MICRO)
- Publisher
- IEEE
- Published
- 2016-10
- Pages
- 1-12
- ISSN
- 1072-4451 (ISSN-L)
- OpenAlex
- W4249932213
- Cited by
- 662
- References
- 20
- Citations by year
- 2026: 20 · 2025: 51 · 2024: 72 · 2023: 67 · 2022: 87 · 2021: 103 · 2020: 72 · 2019: 90 · 2018: 76 · 2017: 23 · 2016: 1
- FWCI
- 21.98
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- IEEE/ACM International Symposium on Microarchitecture (MICRO) (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Hardware and Architecture › Embedded Systems Design Techniques
- Physical Sciences › Computer Science › Hardware and Architecture › Parallel Computing and Optimization Techniques
- Keywords
- CNN accelerator, layer fusion, data flow design, FPGA implementation, convolutional neural network inference
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.