Chaonong Xu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Chaonong Xu has a total of 48 co-authors. Here are the top co-authors by number of shared publications:
Yongjun Xu
Institute of Computing Technology, China Academy of Sciences, Beijing, China
4 shared publications- Yongjun Xu
Chinese Academy of Sciences
4 shared publications Zhizhuo Liu
China University of Petroleum-Beijing
4 shared publicationsZhibang Liu
Beijing Key Lab of Petroleum Data Mining China University of Petroleum Beijing China
3 shared publicationsFan Wu
China University of Petroleum
2 shared publicationsKaichi Ma
2 shared publications
Latest Publications
- 3 January 2026
- Published by Springer Nature
- p. 175-188
A High Throughput Deep Learning Platform for Edge Inference Based on Hardware/Software Codesign
- 3 January 2026
- Published by Springer Nature
- p. 175-188
A High Throughput Deep Learning Platform for Edge Inference Based on Hardware/Software Codesign
- 10 August 2025
- Published by Springer Nature
- p. 83-102
Delay Minimization for Downlink PD-NOMA Transmission with Index Coding in Cache-Aided Wireless Networks
- 10 August 2025
- Published by Springer Nature
- p. 83-102
Delay Minimization for Downlink PD-NOMA Transmission with Index Coding in Cache-Aided Wireless Networks
- 25 July 2025
- Published by Springer Nature
- p. 514-524
FGSMS: Fine-Grained SM Scheduling for Efficient Deep Learning Computing
- 25 July 2025
- Published by Springer Nature
- p. 514-524
FGSMS: Fine-Grained SM Scheduling for Efficient Deep Learning Computing
- Open Access
Collaborative Inference Acceleration with Non-Penetrative Tensor Partitioning
The inference of large-sized images on Internet of Things (IoT) devices is commonly hindered by limited resources, while there are often stringent latency requirements for Deep Neural Network (DNN) inference. Currently, this problem is generally addressed by collaborative inference, where the large-sized image is partitioned into multiple tiles, and each tile is assigned to an IoT device for processing. However, since significant latency will be incurred due to the communication overhead caused by tile sharing, the existing collaborative inference strategy is inefficient for convolutional computation, which is indispensable for any DNN. To reduce it, we propose Non-Penetrative Tensor Partitioning (NPTP), a fine-grained tensor partitioning method that reduces the communication latency by minimizing the communication load of tiles shared, thereby reducing inference latency. We evaluate NPTP with four widely-adopted DNN models. Experimental results demonstrate that NPTP achieves a 1.44-1.68x inference speedup relative to CoEdge, a state-of-the-art (SOTA) collaborative inference algorithm.
Open AccessCollaborative Inference Acceleration with Non-Penetrative Tensor Partitioning
The inference of large-sized images on Internet of Things (IoT) devices is commonly hindered by limited resources, while there are often stringent latency requirements for Deep Neural Network (DNN) inference. Currently, this problem is generally addressed by collaborative inference, where the large-sized image is partitioned into multiple tiles, and each tile is assigned to an IoT device for processing. However, since significant latency will be incurred due to the communication overhead caused by tile sharing, the existing collaborative inference strategy is inefficient for convolutional computation, which is indispensable for any DNN. To reduce it, we propose Non-Penetrative Tensor Partitioning (NPTP), a fine-grained tensor partitioning method that reduces the communication latency by minimizing the communication load of tiles shared, thereby reducing inference latency. We evaluate NPTP with four widely-adopted DNN models. Experimental results demonstrate that NPTP achieves a 1.44-1.68x inference speedup relative to CoEdge, a state-of-the-art (SOTA) collaborative inference algorithm.
- Open Access
Cooperative Inference with Interleaved Operator Partitioning for CNNs
Deploying deep learning models on Internet of Things (IoT) devices often faces challenges due to limited memory resources and computing capabilities. Cooperative inference is an important method for addressing this issue, requiring the partitioning and distributive deployment of an intelligent model. To perform horizontal partitions, existing cooperative inference methods take either the output channel of operators or the height and width of feature maps as the partition dimensions. In this manner, since the activation of operators is distributed, they have to be concatenated together before being fed to the next operator, which incurs the delay for cooperative inference. In this paper, we propose the Interleaved Operator Partitioning (IOP) strategy for CNN models. By partitioning an operator based on the output channel dimension and its successive operator based on the input channel dimension, activation concatenation becomes unnecessary, thereby reducing the number of communication connections, which consequently reduces cooperative inference de-lay. Based on IOP, we further present a model segmentation algorithm for minimizing cooperative inference time, which greedily selects operators for IOP pairing based on the inference delay benefit harvested. Experimental results demonstrate that compared with the state-of-the-art partition approaches used in CoEdge, the IOP strategy achieves 6.39% ~ 16.83% faster acceleration and reduces peak memory footprint by 21.22% ~ 49.98% for three classical image classification models.
Open AccessCooperative Inference with Interleaved Operator Partitioning for CNNs
Deploying deep learning models on Internet of Things (IoT) devices often faces challenges due to limited memory resources and computing capabilities. Cooperative inference is an important method for addressing this issue, requiring the partitioning and distributive deployment of an intelligent model. To perform horizontal partitions, existing cooperative inference methods take either the output channel of operators or the height and width of feature maps as the partition dimensions. In this manner, since the activation of operators is distributed, they have to be concatenated together before being fed to the next operator, which incurs the delay for cooperative inference. In this paper, we propose the Interleaved Operator Partitioning (IOP) strategy for CNN models. By partitioning an operator based on the output channel dimension and its successive operator based on the input channel dimension, activation concatenation becomes unnecessary, thereby reducing the number of communication connections, which consequently reduces cooperative inference de-lay. Based on IOP, we further present a model segmentation algorithm for minimizing cooperative inference time, which greedily selects operators for IOP pairing based on the inference delay benefit harvested. Experimental results demonstrate that compared with the state-of-the-art partition approaches used in CoEdge, the IOP strategy achieves 6.39% ~ 16.83% faster acceleration and reduces peak memory footprint by 21.22% ~ 49.98% for three classical image classification models.
- 14 August 2024
- Published by Springer Nature
- p. 350-361
Efficient Neural Network Fine-Tuning via Layer Contribution Analysis
- 14 August 2024
- Published by Springer Nature
- p. 350-361
Efficient Neural Network Fine-Tuning via Layer Contribution Analysis
- 23 February 2024
- Published by Springer Nature
- p. 173-190
Enhance Broadcasting Throughput by Associating Network Coding with UAVs Relays Deployment in Emergency Communications
- 23 February 2024
- Published by Springer Nature
- p. 173-190
Enhance Broadcasting Throughput by Associating Network Coding with UAVs Relays Deployment in Emergency Communications
- 31 July 2023
- Published by Springer Nature
- p. 15-26
Minimizing Peak Memory Footprint of Inference on IoTs Devices by Efficient Recomputation
- 31 July 2023
- Published by Springer Nature
- p. 15-26
Minimizing Peak Memory Footprint of Inference on IoTs Devices by Efficient Recomputation
- Open Access
- 2 March 2023
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Sensor Networks
- Vol. 19 (3) , 1-22
IMF 2 O 2 : A Fully Connected Sensor Deployment Algorithm for Underwater Sensor Networks
To address the problems of node deployment schemes in existing underwater sensor networks that lack consideration of network connectivity and high deployment costs, this article constructs an optimization model that maximizes network coverage and minimizes deployment costs while ensuring full connectivity. For the NP-hard property of this optimization model, an improved moth flame optimization node deployment algorithm based on fuzzy operators (IMF 2 O 2 ) is proposed. First, comprehensively considering the two performance metrics of network coverage and network connectivity, a multi-objective selection mechanism based on fuzzy operators is proposed to improve network coverage while ensuring full connectivity. Second, a fixed number of nodes are used to monitor the target event points, transforming the node deployment of sensors into an optimal problem and proposing an improved moth flame optimization algorithm to solve this problem. Finally, the two metrics of coverage and deployment cost are measured and the fuzzy operator is used to select the optimal number of nodes to be deployed. Numerical results showed that the proposed algorithm improved network coverage rate by 10%, 22%, and 25%, and improved network connectivity rate by 12%, 20%, and 8% as compared to PSSD, RAWS, and VODA, respectively, while ensuring full connectivity.
- 2 March 2023
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Sensor Networks
- Vol. 19 (3) , 1-22
Open AccessIMF 2 O 2 : A Fully Connected Sensor Deployment Algorithm for Underwater Sensor Networks
To address the problems of node deployment schemes in existing underwater sensor networks that lack consideration of network connectivity and high deployment costs, this article constructs an optimization model that maximizes network coverage and minimizes deployment costs while ensuring full connectivity. For the NP-hard property of this optimization model, an improved moth flame optimization node deployment algorithm based on fuzzy operators (IMF 2 O 2 ) is proposed. First, comprehensively considering the two performance metrics of network coverage and network connectivity, a multi-objective selection mechanism based on fuzzy operators is proposed to improve network coverage while ensuring full connectivity. Second, a fixed number of nodes are used to monitor the target event points, transforming the node deployment of sensors into an optimal problem and proposing an improved moth flame optimization algorithm to solve this problem. Finally, the two metrics of coverage and deployment cost are measured and the fuzzy operator is used to select the optimal number of nodes to be deployed. Numerical results showed that the proposed algorithm improved network coverage rate by 10%, 22%, and 25%, and improved network connectivity rate by 12%, 20%, and 8% as compared to PSSD, RAWS, and VODA, respectively, while ensuring full connectivity.
- 2 February 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Systems Journal
- Vol. 16 (1) , 810-819
A Low-Latency Random Access Scheme by Multichannel SIC for Industrial IoT
Low latency is a prerequisite for real-time applications, and random access is almost an inevitable choice for large-scale wireless networks. In this article, we consider how to decrease access delay in random access manners for the industrial Internet of Things. We propose a novel scheme by integrating multiple subchannels on the user equipment (UE) side with message-level successive interference cancellation (SIC) on the sink side. Specifically, for any UE that starts an uplink transmission, it just stochastically chooses parts of its subchannels and then simultaneously transmits the same message on them without any power control. On the sink side, the sink decodes signals from UEs using SIC across multiple subchannels. Based on the proposed multichannel SIC scheme, minimizing the access delay in the random access manner is equivalent to minimizing the probability of decoding deadlock. Based on the mathematical method of density evolution and the help from the differential evolution algorithm, an optimal distribution for choosing subchannels is presented. With the optimal distribution, the average access delay is minimized.
- 2 February 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Systems Journal
- Vol. 16 (1) , 810-819
A Low-Latency Random Access Scheme by Multichannel SIC for Industrial IoT
Low latency is a prerequisite for real-time applications, and random access is almost an inevitable choice for large-scale wireless networks. In this article, we consider how to decrease access delay in random access manners for the industrial Internet of Things. We propose a novel scheme by integrating multiple subchannels on the user equipment (UE) side with message-level successive interference cancellation (SIC) on the sink side. Specifically, for any UE that starts an uplink transmission, it just stochastically chooses parts of its subchannels and then simultaneously transmits the same message on them without any power control. On the sink side, the sink decodes signals from UEs using SIC across multiple subchannels. Based on the proposed multichannel SIC scheme, minimizing the access delay in the random access manner is equivalent to minimizing the probability of decoding deadlock. Based on the mathematical method of density evolution and the help from the differential evolution algorithm, an optimal distribution for choosing subchannels is presented. With the optimal distribution, the average access delay is minimized.