Jiawei Nian
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Jiawei Nian has a total of 48 co-authors. Here are the top co-authors by number of shared publications:
- Fang Fang
North China Electric Power University
6 shared publications - Long Cheng
North China Electric Power University
5 shared publications - Mengfei Yang
China Academy of Space Technology
4 shared publications - Xin Gao
Beijing SunWise Space Technology Ltd, beijing China
4 shared publications - Xingtang Wu
School of Control and Computer Engineering, North China Electric Power University, Beijing, China
4 shared publications - Cong Liu
Nova Information Management School, Universidade Nova de Lisboa, Lisbon, Portugal
2 shared publications
Latest Publications
- Open Access
- 13 May 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Embedded Computing Systems
- Vol. 25 (3) , 1-20
MCU-MixQ: A HW/SW Co-optimized Mixed-precision Neural Network Design Framework for MCUs
Mixed-precision neural network (MPNN) that utilizes just enough data width for the neural network processing is an effective approach to meet the stringent resources constraints including memory and computing of MCUs. Nevertheless, there is still a lack of sub-byte and mixed-precision SIMD operations in MCU-class ISA and the limited computing capability of MCUs remains underutilized, which further aggravates the computing bound encountered in neural network processing. As a result, the benefits of MPNNs cannot be fully unleashed. In this work, we propose to pack multiple low-bitwidth arithmetic operations within a single instruction multiple data (SIMD) instructions in typical MCUs, and then develop an efficient convolution operator by exploring both the data parallelism and computing parallelism in convolution along with the proposed SIMD packing. Finally, we further leverage Neural Architecture Search (NAS) to build a HW/SW co-designed MPNN design framework, namely MCU-MixQ. This framework can optimize both the MPNN quantization and MPNN implementation efficiency, striking an optimized balance between neural network performance and accuracy. According to our experiment results, MCU-MixQ achieves 2.1× and 1.4× speedup over CMix-NN and MCUNet respectively under the same resource constraints. MCU-MixQ is also open sourced on GitHub. 1
- 13 May 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Embedded Computing Systems
- Vol. 25 (3) , 1-20
Open AccessMCU-MixQ: A HW/SW Co-optimized Mixed-precision Neural Network Design Framework for MCUs
Mixed-precision neural network (MPNN) that utilizes just enough data width for the neural network processing is an effective approach to meet the stringent resources constraints including memory and computing of MCUs. Nevertheless, there is still a lack of sub-byte and mixed-precision SIMD operations in MCU-class ISA and the limited computing capability of MCUs remains underutilized, which further aggravates the computing bound encountered in neural network processing. As a result, the benefits of MPNNs cannot be fully unleashed. In this work, we propose to pack multiple low-bitwidth arithmetic operations within a single instruction multiple data (SIMD) instructions in typical MCUs, and then develop an efficient convolution operator by exploring both the data parallelism and computing parallelism in convolution along with the proposed SIMD packing. Finally, we further leverage Neural Architecture Search (NAS) to build a HW/SW co-designed MPNN design framework, namely MCU-MixQ. This framework can optimize both the MPNN quantization and MPNN implementation efficiency, striking an optimized balance between neural network performance and accuracy. According to our experiment results, MCU-MixQ achieves 2.1× and 1.4× speedup over CMix-NN and MCUNet respectively under the same resource constraints. MCU-MixQ is also open sourced on GitHub. 1
- 29 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computers
- Vol. 75 (8) , 2797-2810
Prototype Retrieval-Augmented Federated Learning System for Robust Intrusion Detection
Detecting malicious attacks is essential for protecting computer systems and ensuring device security. FederatedLearning (FL)-based Intrusion Detection Systems (IDS) haveemerged as promising solutions, enabling multiple clients (i.e.,data owners) to collaboratively train intrusion detection modelswithout sharing private data. However, current FL studies typicallyassume that each client’s training and test label distributionsare identical. This assumption is overly idealistic and rarelyholds in real-world scenarios, leading to suboptimal performancewhen label distribution shifts occur between the training andtesting data. To address this challenge, we propose FedPRO, a plug-and-play framework designed to improve the test-timeperformance of existing FL methods, without modifying theiroriginal training pipelines or fine-tuning the trained FL models.Specifically, we develop a unique prototype generation andoptimization mechanism to produce semantically meaningfulclass prototypes. These prototypes constitute a prototype memorybank, serving as an external knowledge repository. At test time,a prototype retrieval-augmented inference strategy is employedto query relevant prototypes and refine predictions on eachclient, effectively alleviating the label distribution shift issues andboosting prediction accuracy.We evaluate FedPRO by integratingit with various off-the-shelf FL methods. Extensive experimentalresults on benchmark datasets demonstrate its effectiveness andscalability. Notably, applying FedPRO to the state-of-the-artmethod FedDBE improves its test accuracy from 79.25% to86.66% on the CICIDS-2018 dataset, while introducing onlyapproximately 32KB of additional communication overhead. Ourcode is available at: https://github.com/zza234s/FedPRO
- 29 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computers
- Vol. 75 (8) , 2797-2810
Prototype Retrieval-Augmented Federated Learning System for Robust Intrusion Detection
Detecting malicious attacks is essential for protecting computer systems and ensuring device security. FederatedLearning (FL)-based Intrusion Detection Systems (IDS) haveemerged as promising solutions, enabling multiple clients (i.e.,data owners) to collaboratively train intrusion detection modelswithout sharing private data. However, current FL studies typicallyassume that each client’s training and test label distributionsare identical. This assumption is overly idealistic and rarelyholds in real-world scenarios, leading to suboptimal performancewhen label distribution shifts occur between the training andtesting data. To address this challenge, we propose FedPRO, a plug-and-play framework designed to improve the test-timeperformance of existing FL methods, without modifying theiroriginal training pipelines or fine-tuning the trained FL models.Specifically, we develop a unique prototype generation andoptimization mechanism to produce semantically meaningfulclass prototypes. These prototypes constitute a prototype memorybank, serving as an external knowledge repository. At test time,a prototype retrieval-augmented inference strategy is employedto query relevant prototypes and refine predictions on eachclient, effectively alleviating the label distribution shift issues andboosting prediction accuracy.We evaluate FedPRO by integratingit with various off-the-shelf FL methods. Extensive experimentalresults on benchmark datasets demonstrate its effectiveness andscalability. Notably, applying FedPRO to the state-of-the-artmethod FedDBE improves its test accuracy from 79.25% to86.66% on the CICIDS-2018 dataset, while introducing onlyapproximately 32KB of additional communication overhead. Ourcode is available at: https://github.com/zza234s/FedPRO
- 23 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 75 (00189456) , 1-15
A Few-Shot Wind Turbine Blade Surface Faults Detection via Enhanced Data Augmentation and Progressive Transfer Learning
Accurate and automated measurement of surface faults on wind turbine blades (WTBs) is essential for ensuring the structural integrity and operational safety of wind farms. Surface faults on WTBs not only shorten turbine lifespans but may also compromise the stability and accuracy of measurement and monitoring instruments in wind farms. Traditional measurement approaches demonstrated promising performance for WTB surface faults with a large amount of fault data, but acquiring such extensive data is often impractical in real industry situations. To overcome data scarcity and accuracy loss in fault detection, we propose an image augmentation and detection with adaptive learning framework (IADA-Net), consisting of a WTB image augmentation module and a surface fault detection module. The real blade image augmentation module designs an augmentation algorithm (HybridAug), which expands the blade faults dataset across varying acquisition scenarios and environments. The surface fault detection module (TransDetect) designs a detection algorithm based on transfer learning to address limited model generalization and diminished detection accuracy in few-shot scenarios. We conducted experiments on 1223 real images collected from China Longyuan Power’s wind farms. The experimental results show that the proposed HybridAug and TransDetect can improve the WTB fault detection accuracy by 2.4% and 18.6%, respectively. The IADA-Net achieves a detection accuracy of 96.7%, comparable to YOLO-based models, while reducing the training data requirement by 60%, which demonstrates our method offers a reliable solution for intelligent blade operation and maintenance systems.
- 23 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 75 (00189456) , 1-15
A Few-Shot Wind Turbine Blade Surface Faults Detection via Enhanced Data Augmentation and Progressive Transfer Learning
Accurate and automated measurement of surface faults on wind turbine blades (WTBs) is essential for ensuring the structural integrity and operational safety of wind farms. Surface faults on WTBs not only shorten turbine lifespans but may also compromise the stability and accuracy of measurement and monitoring instruments in wind farms. Traditional measurement approaches demonstrated promising performance for WTB surface faults with a large amount of fault data, but acquiring such extensive data is often impractical in real industry situations. To overcome data scarcity and accuracy loss in fault detection, we propose an image augmentation and detection with adaptive learning framework (IADA-Net), consisting of a WTB image augmentation module and a surface fault detection module. The real blade image augmentation module designs an augmentation algorithm (HybridAug), which expands the blade faults dataset across varying acquisition scenarios and environments. The surface fault detection module (TransDetect) designs a detection algorithm based on transfer learning to address limited model generalization and diminished detection accuracy in few-shot scenarios. We conducted experiments on 1223 real images collected from China Longyuan Power’s wind farms. The experimental results show that the proposed HybridAug and TransDetect can improve the WTB fault detection accuracy by 2.4% and 18.6%, respectively. The IADA-Net achieves a detection accuracy of 96.7%, comparable to YOLO-based models, while reducing the training data requirement by 60%, which demonstrates our method offers a reliable solution for intelligent blade operation and maintenance systems.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 75, 3526515
A Distributed Acoustic Measurement and Diagnosis System of Wind Turbine Blades
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 75, 3526515
A Distributed Acoustic Measurement and Diagnosis System of Wind Turbine Blades
- 21 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 21 (11) , 8305-8314
Spatio-Temporal Feature Extraction for Predicting Large-Scale Train Delay Propagation in High-Speed Railway Networks
The safe, punctual, and reliable operation of high-speed railway (HSR) networks is crucial for ensuring system efficiency and enhancing passenger experience. However, due to the complexity of HSR systems and long operation routes, failures in system components can lead to unexpected incidents, causing deviations from the scheduled operations and, subsequently, delays. These delays, particularly large delays, significantly impact the overall performance and passenger experience of the network. Thus, accurate prediction of train delays, especially in large-delay scenarios, is essential for optimizing train scheduling and restoring normal operations in a timely manner. To address this challenge, this study proposes a deep learning architecture based on spatio-temporal feature extraction for accurate train delay prediction. A graph attention network-long short-term memory block is introduced to capture the spatio-temporal evolution features of different trains. In addition, a sequence forecasting approach and a mixture of experts module are integrated to model the complex relationships between the target train’s delay and its previous states. The delay evolution features are incorporated into the loss function, allowing for more accurate predictions. Experimental results show that the proposed model outperforms the baseline models, achieving at least an improvement of 50.62% and 27.89% in root-mean-squared error and mean absolute error, respectively. When the error tolerance is set within 3 min, the prediction accuracy reaches 96.68%. Experimental results demonstrate the superior performance of the proposed model.
- 21 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 21 (11) , 8305-8314
Spatio-Temporal Feature Extraction for Predicting Large-Scale Train Delay Propagation in High-Speed Railway Networks
The safe, punctual, and reliable operation of high-speed railway (HSR) networks is crucial for ensuring system efficiency and enhancing passenger experience. However, due to the complexity of HSR systems and long operation routes, failures in system components can lead to unexpected incidents, causing deviations from the scheduled operations and, subsequently, delays. These delays, particularly large delays, significantly impact the overall performance and passenger experience of the network. Thus, accurate prediction of train delays, especially in large-delay scenarios, is essential for optimizing train scheduling and restoring normal operations in a timely manner. To address this challenge, this study proposes a deep learning architecture based on spatio-temporal feature extraction for accurate train delay prediction. A graph attention network-long short-term memory block is introduced to capture the spatio-temporal evolution features of different trains. In addition, a sequence forecasting approach and a mixture of experts module are integrated to model the complex relationships between the target train’s delay and its previous states. The delay evolution features are incorporated into the loss function, allowing for more accurate predictions. Experimental results show that the proposed model outperforms the baseline models, achieving at least an improvement of 50.62% and 27.89% in root-mean-squared error and mean absolute error, respectively. When the error tolerance is set within 3 min, the prediction accuracy reaches 96.68%. Experimental results demonstrate the superior performance of the proposed model.
- 27 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Systems Journal
- Vol. 19 (2) , 471-482
Quantum Reinforcement Learning for QoS-Aware Real-Time Job Scheduling in Cloud Systems
Effective cloud job scheduling is essential for enhancing the performance and operational efficiency of cloud-based services, directly impacting their quality of service (QoS). Among existing methodologies, deep reinforcement learning (DRL) has proven effective in addressing complex, multidimensional optimization challenges in real-time scheduling. With advancements in quantum computing, quantum neural networks (QNNs) are showing unique advantages in information representation and processing. This study is the first to explore quantum reinforcement learning (QRL) for real-time job scheduling in cloud systems. Specifically, we propose a QRL framework that utilizes variational and encoding layers to convert state information into quantum data, repeatedly embedded into a QNN to compute optimal value returns. This approach aims to enhance QoS by improving job execution success rates and reducing average response times with unpredictable job arrivals. We present the detailed design of our approach, and our simulation results demonstrate that the QRL method significantly exceeds established baselines, including those based on DRL, across a range of workload intensities and computational resource configurations. This is particularly evident under high-load conditions, where our approach can achieve 55.2% higher success rates, underscoring its significant potential in cloud job scheduling optimization.
- 27 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Systems Journal
- Vol. 19 (2) , 471-482
Quantum Reinforcement Learning for QoS-Aware Real-Time Job Scheduling in Cloud Systems
Effective cloud job scheduling is essential for enhancing the performance and operational efficiency of cloud-based services, directly impacting their quality of service (QoS). Among existing methodologies, deep reinforcement learning (DRL) has proven effective in addressing complex, multidimensional optimization challenges in real-time scheduling. With advancements in quantum computing, quantum neural networks (QNNs) are showing unique advantages in information representation and processing. This study is the first to explore quantum reinforcement learning (QRL) for real-time job scheduling in cloud systems. Specifically, we propose a QRL framework that utilizes variational and encoding layers to convert state information into quantum data, repeatedly embedded into a QNN to compute optimal value returns. This approach aims to enhance QoS by improving job execution success rates and reducing average response times with unpredictable job arrivals. We present the detailed design of our approach, and our simulation results demonstrate that the QRL method significantly exceeds established baselines, including those based on DRL, across a range of workload intensities and computational resource configurations. This is particularly evident under high-load conditions, where our approach can achieve 55.2% higher success rates, underscoring its significant potential in cloud job scheduling optimization.
- Open Access
- 7 April 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Embedded Computing Systems
RPFF-PA : Reliable and Parallel Fault-tolerant Framework for Path Latency Reduction Deployed in Register Arrays
Soft errors can lead to electronic device failures caused by radiation events. Therefore, fault-tolerant designs are essential for safeguarding critical memory components and processors. The N-modular redundancy (NMR) scheme presents a practical means of implementing register protection. However, the NMR fault-tolerant scheme is serially configured on the critical path, increasing path latency of approximately 12.4 \(\% \) -35.5 \(\% \) . In this paper, we propose a parallel fault-tolerant framework for register arrays(RPFF-PA), which realizes the parallelized execution of hardern processes and read/write operations. By executing the decoding process in parallel with the decision circuit, the latency cost caused by fault-tolerant components is significantly reduced. The results demonstrate that the strategy reduces path latency by 5.2 \(\% \) -20.3 \(\% \) compared to the original NMR approach. Furthermore, the framework includes a real-time error correction module capable of effectively addressing the error accumulation issue, thereby enhancing the reliability of the fault-tolerant system.- 7 April 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Embedded Computing Systems
Open AccessRPFF-PA : Reliable and Parallel Fault-tolerant Framework for Path Latency Reduction Deployed in Register Arrays
Soft errors can lead to electronic device failures caused by radiation events. Therefore, fault-tolerant designs are essential for safeguarding critical memory components and processors. The N-modular redundancy (NMR) scheme presents a practical means of implementing register protection. However, the NMR fault-tolerant scheme is serially configured on the critical path, increasing path latency of approximately 12.4 \(\% \) -35.5 \(\% \) . In this paper, we propose a parallel fault-tolerant framework for register arrays(RPFF-PA), which realizes the parallelized execution of hardern processes and read/write operations. By executing the decoding process in parallel with the decision circuit, the latency cost caused by fault-tolerant components is significantly reduced. The results demonstrate that the strategy reduces path latency by 5.2 \(\% \) -20.3 \(\% \) compared to the original NMR approach. Furthermore, the framework includes a real-time error correction module capable of effectively addressing the error accumulation issue, thereby enhancing the reliability of the fault-tolerant system. - Open Access
- 1 April 2024
- Published by MDPI AG in Electronics
- Vol. 13 (7) , 1320
Two-Dimensional Protection Code for Virtual Page Information in Translation Lookaside Buffers
Severe conditions such as high-energy particle strikes may induce soft errors in on-chip memory, like cache and translation lookaside buffers (TLBs). As the key component of virtual-to-physical address translation, TLB directly affects processor performance. To protect the virtual page information stored in TLB, several studies have introduced error detection or correction codes. However, most schemes proposed for data TLB cannot support the detection of multi-bit upsets, which is reported as a serious issue in modern fault-tolerant processors due to the downscaling of CMOS process technology. In this paper, we propose a new two-dimensional protection technique, called the matrix protection tag (MaP-Tag), to provide stronger error detection and correction capability to TLB virtual page information. Our proposal reorganizes virtual page information into a matrix and employs adjacent check bits and interleaved check bits for error detection. The two-dimensional design offers one-bit error detection, burst multi-bit error detection, and even multi-bit error correction in some cases. The simulation results show that our proposal can detect almost all error patterns when injected with up to eight bit-flips. Furthermore, the technique provides a better error correction rate than conventional single error correction (SEC) codes. The reliability calculation shows that our proposal is powerful in both error detection and correction with affordable storage overhead.
- 1 April 2024
- Published by MDPI AG in Electronics
- Vol. 13 (7) , 1320
Open AccessTwo-Dimensional Protection Code for Virtual Page Information in Translation Lookaside Buffers
Severe conditions such as high-energy particle strikes may induce soft errors in on-chip memory, like cache and translation lookaside buffers (TLBs). As the key component of virtual-to-physical address translation, TLB directly affects processor performance. To protect the virtual page information stored in TLB, several studies have introduced error detection or correction codes. However, most schemes proposed for data TLB cannot support the detection of multi-bit upsets, which is reported as a serious issue in modern fault-tolerant processors due to the downscaling of CMOS process technology. In this paper, we propose a new two-dimensional protection technique, called the matrix protection tag (MaP-Tag), to provide stronger error detection and correction capability to TLB virtual page information. Our proposal reorganizes virtual page information into a matrix and employs adjacent check bits and interleaved check bits for error detection. The two-dimensional design offers one-bit error detection, burst multi-bit error detection, and even multi-bit error correction in some cases. The simulation results show that our proposal can detect almost all error patterns when injected with up to eight bit-flips. Furthermore, the technique provides a better error correction rate than conventional single error correction (SEC) codes. The reliability calculation shows that our proposal is powerful in both error detection and correction with affordable storage overhead.