Changqing Xu

Prof. Changqing Xu

Xidian University, Xi'an, China
orcid

Research Focus & Metrics

Academic Career Length

10Years (2017-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 October 2021 in IEEE Transactions on Electron Devices
Verified
Journal Article
Closed access
Changqing XuChangqing Xu
Changqing Xu
YLYi Liu
Yi Liu
XLXinfang Liao
Xinfang Liao
JCJialiang Cheng
Jialiang Cheng
YYYintang Yang
Yintang Yang
In this article, a novel machine learning regression-based single-event transient (SET) modeling method is proposed. The proposed method can obtain a reasonable and accurate SET model without introducing complex physical mechanisms, which are not suitable for circuit-level simulation, into the model. To capture the essential physics behind these current transients caused by SET in the circuit-level simulations, we collect plenty of SET current data under different conditions [e.g., different linear energy transfer (LET), different drain bias voltage, different strike position, etc.] to train the SET model. To show the effectiveness of the proposed modeling method, we build a SET pulse current model by learning SET current data of Semiconductor Manufacturing International Corporation (SMIC) 130-nm bulk CMOS obtained by TCAD simulation. A multilayer feedforward neural network is used to build the SET pulse current model and the built SET model takes into account the dependence of time, LETs, drain bias voltages, and strike positions. The results from the model are validated with the simulation from TCAD. The trained SET pulse current model is implemented as a Verilog-A current source in the Cadence Specter circuit simulator, and an inverter with five fan-outs is used to show the practicability and reasonableness of the proposed SET pulse current model for circuit-level single-event effect (SEE) simulation.

19 May 2022 in IEEE Embedded Systems Letters
Verified
Journal Article
Closed access
YLYi Liu
Yi Liu
RGRujia Guo
Rujia Guo
Changqing XuChangqing Xu
Changqing Xu
XWXiaodong Weng
Xiaodong Weng
Yintang YangYintang Yang
Yintang Yang
Link failure and network congestion have a serious impact on the performance degradation of network on chip (NoC). Adaptive routing algorithms, which have the capability of fault tolerance and congestion awareness, can make NoCs adapt to complex and changeable applications. The proposed algorithm in this letter is an adaptive routing algorithm based on $Q$ -learning. By introducing $H - Q$ table to learn the congestion and fault information of the paths between the source node and the destination node, it can avoid the unnecessary detour of data in the fault area and select the path with less congestion. Aiming at the problem caused by the untimely updating of some values in $Q$ -learning, the $Q$ -value decay and dynamic learning rate mechanism are proposed. Experimental results demonstrate that the proposed routing algorithms still can obtain a good performance under a high failure rate (Fault rate >25%) with modest extra hardware overhead.

15 March 2018 in IET Computers & Digital Techniques
Verified
Journal Article
Open access
PDF file
Changqing XuChangqing Xu
Changqing Xu
YLYi Liu
Yi Liu
PLPeng Li
Peng Li
YYYinTang Yang
YinTang Yang
In this study, a flexible energy‐ and delay‐aware mapping approach is proposed for the co‐optimisation of energy consumption and communication latency for network‐on‐chips (NoCs). A novel genetic‐based hyper‐heuristic algorithm (GHA) is proposed as the core algorithm. This algorithm consists of bottom‐level optimisation which includes a variety of operators and top‐level optimisation which selects suitable operators through a ‘reward’ mechanism. As this algorithm can select suitable operators automatically during the mapping process, it noticeably improves convergence speed and demonstrates excellent stability. Compared to the random algorithm, GHA can achieve on average 23.28% delay reduction and 11.81% power reduction. Compared to state‐of‐the‐art mapping algorithms, GHA produces improved mapping results with less time, especially when the size of NoC is large.

13 September 2023 in Frontiers in Neuroscience
Verified
Journal Article
Open access
YPYijian Pei
Yijian Pei
Changqing XuChangqing Xu
Changqing Xu
ZWZili Wu
Zili Wu
YLYi Liu
Yi Liu
YYYintang Yang
Yintang Yang
Spiking neural network (SNN) is a brain-inspired model with more spatio-temporal information processing capacity and computational energy efficiency. However, with the increasing depth of SNNs, the memory problem caused by the weights of SNNs has gradually attracted attention. In this study, we propose an ultra-low latency adaptive local binary spiking neural network (ALBSNN) with accuracy loss estimators, which dynamically selects the network layers to be binarized to ensure a balance between quantization degree and classification accuracy by evaluating the error caused by the binarized weights during the network learning process. At the same time, to accelerate the training speed of the network, the global average pooling (GAP) layer is introduced to replace the fully connected layers by combining convolution and pooling. Finally, to further reduce the error caused by the binary weight, we propose binary weight optimization (BWO), which updates the overall weight by directly adjusting the binary weight. This method further reduces the loss of the network that reaches the training bottleneck. The combination of the above methods balances the network's quantization and recognition ability, enabling the network to maintain the recognition capability equivalent to the full precision network and reduce the storage space by more than 20%. So, SNNs can use a small number of time steps to obtain better recognition accuracy. In the extreme case of using only a one-time step, we still can achieve 93.39, 92.12, and 69.55% testing accuracy on three traditional static datasets, Fashion- MNIST, CIFAR-10, and CIFAR-100, respectively. At the same time, we evaluate our method on neuromorphic N-MNIST, CIFAR10-DVS, and IBM DVS128 Gesture datasets and achieve advanced accuracy in SNN with binary weights. Our network has greater advantages in terms of storage resources and training time.

14 February 2020 in Frontiers in Neuroscience
Verified
Journal Article
Open access
PDF file
Changqing XuChangqing Xu
Changqing Xu
WZWenrui Zhang
Wenrui Zhang
YLYu Liu
Yu Liu
PLPeng Li
Peng Li
Spiking neural networks (SNNs) are the third generation of neural networks and can explore both rate and temporal coding for energy-efficient event-driven computation. However, the decision accuracy of existing SNN designs is contingent upon processing a large number of spikes over a long period. Nevertheless, the switching power of SNN hardware accelerators is proportional to the number of spikes processed while the length of spike trains limits throughput and static power efficiency. This paper presents the first study on developing temporal compression to significantly boost throughput and reduce energy dissipation of digital hardware SNN accelerators while being applicable to multiple spike codes. The proposed compression architectures consist of low-cost input spike compression units, novel input-and-output-weighted spiking neurons, and reconfigurable time constant scaling to support large and flexible time compression ratios. Our compression architectures can be transparently applied to any given pre-designed SNNs employing either rate or temporal codes while incurring minimal modification of the neural models, learning algorithms, and hardware design. Using spiking speech and image recognition datasets, we demonstrate the feasibility of supporting large time compression ratios of up to 16×, delivering up to 15.93×, 13.88×, and 86.21× improvements in throughput, energy dissipation, the tradeoffs between hardware area, runtime, energy, and classification accuracy, respectively based on different spike codes on a Xilinx Zynq-7000 FPGA. These results are achieved while incurring little extra hardware overhead.

4 January 2022 in IEEE Transactions on Electron Devices
Verified
Journal Article
Closed access
XLXinfang Liao
Xinfang Liao
YLYi Liu
Yi Liu
Changqing XuChangqing Xu
Changqing Xu
JLJing Li
Jing Li
YYYintang Yang
Yintang Yang
High-voltage p-i-n diodes for space applications are sensitive to single event burnout (SEB), which may cause the failure of the entire electronic system. This article describes the SEB failure mechanism of high-voltage p-i-n diodes with field limiting rings (FLRs) termination structure, based on the electro-thermal coupled transient simulation model in the Sentaurus TCAD simulator. The simulation results show that the catastrophic failures result from the intense local heating at the edge of the main junction, because the high electric field strength at this region induces strong impact ionization of the charge carriers. The strike at the edge of the main junction shows the strongest sensitivity to SEB since the energetic particle-induced carriers are deposited exactly at the high electric field region. Then, a novel SEB hardening technique of localized carrier lifetime control at the edge of the main junction region is proposed based on the analysis on the failure mechanism. Our simulation results demonstrate that the reduction of the localized carrier lifetime can alleviate the peak electric field and reduce the local heat accumulation, thereby improving the SEB performance of high-voltage p-i-n diodes.