Yanjun Zhang
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
Yanjun Zhang has a total of 50 co-authors. Here are the top co-authors by number of shared publications:
- Bo Yu
Shenzhen Institute of Artificial Intelligence and Robotics for Society, Shenzhen, Guangdong, China
4 shared publications - Hongzheng Tian
University of California Irvine
3 shared publications - Sitao Huang
University of California, Irvine, Irvine, CA, USA
3 shared publications - Xiaoyu Niu
Beijing Institute of Technology
3 shared publications - Yifan Zhang
University of California, Irvine, Irvine, CA, USA
3 shared publications - Huiqi Li
Beijing Institute of Technology
2 shared publications
Latest Publications
- Open Access
- 7 January 2026
- Published by MDPI AG in Biomimetics
- Vol. 11 (1) , 47
Bio-Inspired Neural Network Dynamics-Aware Reinforcement Learning for Spiking Neural Network
Artificial Intelligence (AI) has seen rapid advancements in recent times, finding applications across various sectors and achieving notable successes. However, current AI models based on Deep Convolutional Neural Networks (DNNs) face numerous challenges, particularly a lack of interpretability, which severely restricts their future potential. Spiking Neural Networks (SNNs), considered the third generation of Artificial Neural Networks (ANNs), are at the forefront of brain-inspired AI research. The resemblance between SNNs and biological neural networks offers the potential to create more human-like AI systems with enhanced interpretability, paving the way for more trustworthy AI implementations. Despite this promise, the absence of efficient training methods for large-scale and complex SNNs hampers their broader application. This paper investigates bio-inspired reinforcement learning strategies by examining neural network dynamics during SNN training. The aim is to improve learning efficiency and effectiveness for extensive and intricate SNNs. Our findings suggest that using reinforcement learning to focus on neural network dynamics may be a promising approach for developing learning algorithms for future large-scale SNNs.
- 7 January 2026
- Published by MDPI AG in Biomimetics
- Vol. 11 (1) , 47
Open AccessBio-Inspired Neural Network Dynamics-Aware Reinforcement Learning for Spiking Neural Network
Artificial Intelligence (AI) has seen rapid advancements in recent times, finding applications across various sectors and achieving notable successes. However, current AI models based on Deep Convolutional Neural Networks (DNNs) face numerous challenges, particularly a lack of interpretability, which severely restricts their future potential. Spiking Neural Networks (SNNs), considered the third generation of Artificial Neural Networks (ANNs), are at the forefront of brain-inspired AI research. The resemblance between SNNs and biological neural networks offers the potential to create more human-like AI systems with enhanced interpretability, paving the way for more trustworthy AI implementations. Despite this promise, the absence of efficient training methods for large-scale and complex SNNs hampers their broader application. This paper investigates bio-inspired reinforcement learning strategies by examining neural network dynamics during SNN training. The aim is to improve learning efficiency and effectiveness for extensive and intricate SNNs. Our findings suggest that using reinforcement learning to focus on neural network dynamics may be a promising approach for developing learning algorithms for future large-scale SNNs.
- Open Access
- 24 July 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
A Sparsity-Aware Autonomous Path Planning Accelerator with HW/SW Co-Design and Multi-Level Dataflow Optimization
Path planning is a critical task for autonomous driving, aiming to generate smooth, collision-free, and feasible paths based on input perception and localization information. The planning task is both highly time-sensitive and computationally intensive, posing significant challenges to resource-constrained autonomous driving hardware. In this paper, we propose an end-to-end framework for accelerating path planning on FPGA platforms. This framework focuses on accelerating quadratic programming (QP) solving, which is the core of optimization-based path planning and has the most computationally-intensive workloads. Our method leverages a hardware-friendly alternating direction method of multipliers (ADMM) to solve QP problems while employing a highly parallelizable preconditioned conjugate gradient (PCG) method for solving the associated linear systems. We analyze the sparse patterns of matrix operations in QP and design customized storage schemes along with efficient sparse matrix multiplication and sparse matrix-vector multiplication units. Our customized design significantly reduces resource consumption for data storage and computation while dramatically speeding up matrix operations. Additionally, we propose a multi-level dataflow optimization strategy. Within individual operators, we achieve acceleration through parallelization and pipelining. For different operators in an algorithm, we analyze inter-operator data dependencies to enable fine-grained pipelining. At the system level, we map different steps of the planning process to the CPU and FPGA and pipeline these steps to enhance end-to-end throughput. We implement and validate our design on the AMD ZCU102 platform. Our implementation achieves state-of-the-art performance in both latency and energy efficiency compared to existing works, including an average 1.48 × speedup over the best FPGA-based design, a 2.89 × speedup compared to the state-of-the-art QP solver on an Intel i7-11800H CPU, a 5.62 × speedup over an ARM Cortex-A57 embedded CPU, and a 1.56 × speedup over state-of-the-art GPU-based work. Furthermore, our design delivers a 2.05 × improvement in throughput compared to the state-of-the-art FPGA-based design.
- 24 July 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
Open AccessA Sparsity-Aware Autonomous Path Planning Accelerator with HW/SW Co-Design and Multi-Level Dataflow Optimization
Path planning is a critical task for autonomous driving, aiming to generate smooth, collision-free, and feasible paths based on input perception and localization information. The planning task is both highly time-sensitive and computationally intensive, posing significant challenges to resource-constrained autonomous driving hardware. In this paper, we propose an end-to-end framework for accelerating path planning on FPGA platforms. This framework focuses on accelerating quadratic programming (QP) solving, which is the core of optimization-based path planning and has the most computationally-intensive workloads. Our method leverages a hardware-friendly alternating direction method of multipliers (ADMM) to solve QP problems while employing a highly parallelizable preconditioned conjugate gradient (PCG) method for solving the associated linear systems. We analyze the sparse patterns of matrix operations in QP and design customized storage schemes along with efficient sparse matrix multiplication and sparse matrix-vector multiplication units. Our customized design significantly reduces resource consumption for data storage and computation while dramatically speeding up matrix operations. Additionally, we propose a multi-level dataflow optimization strategy. Within individual operators, we achieve acceleration through parallelization and pipelining. For different operators in an algorithm, we analyze inter-operator data dependencies to enable fine-grained pipelining. At the system level, we map different steps of the planning process to the CPU and FPGA and pipeline these steps to enhance end-to-end throughput. We implement and validate our design on the AMD ZCU102 platform. Our implementation achieves state-of-the-art performance in both latency and energy efficiency compared to existing works, including an average 1.48 × speedup over the best FPGA-based design, a 2.89 × speedup compared to the state-of-the-art QP solver on an Intel i7-11800H CPU, a 5.62 × speedup over an ARM Cortex-A57 embedded CPU, and a 1.56 × speedup over state-of-the-art GPU-based work. Furthermore, our design delivers a 2.05 × improvement in throughput compared to the state-of-the-art FPGA-based design.
- Open Access
- 14 January 2025
- Published by MDPI AG in Biomimetics
- Vol. 10 (1) , 48
Biologically Inspired Spatial–Temporal Perceiving Strategies for Spiking Neural Network
A future unmanned system needs the ability to perceive, decide and control in an open dynamic environment. In order to fulfill this requirement, it needs to construct a method with a universal environmental perception ability. Moreover, this perceptual process needs to be interpretable and understandable, so that future interactions between unmanned systems and humans can be unimpeded. However, current mainstream DNN (deep learning neural network)-based AI (artificial intelligence) is a ‘black box’. We cannot interpret or understand how the decision is made by these AIs. An SNN (spiking neural network), which is more similar to a biological brain than a DNN, has the potential to implement interpretable or understandable AI. In this work, we propose a neuron group-based structural learning method for an SNN to better capture the spatial and temporal information from the external environment, and propose a time-slicing scheme to better interpret the spatial and temporal information of responses generated by an SNN. Results show that our method indeed helps to enhance the environment perception ability of the SNN, and possesses a certain degree of robustness, enhancing the potential to build an interpretable or understandable AI in the future.
- 14 January 2025
- Published by MDPI AG in Biomimetics
- Vol. 10 (1) , 48
Open AccessBiologically Inspired Spatial–Temporal Perceiving Strategies for Spiking Neural Network
A future unmanned system needs the ability to perceive, decide and control in an open dynamic environment. In order to fulfill this requirement, it needs to construct a method with a universal environmental perception ability. Moreover, this perceptual process needs to be interpretable and understandable, so that future interactions between unmanned systems and humans can be unimpeded. However, current mainstream DNN (deep learning neural network)-based AI (artificial intelligence) is a ‘black box’. We cannot interpret or understand how the decision is made by these AIs. An SNN (spiking neural network), which is more similar to a biological brain than a DNN, has the potential to implement interpretable or understandable AI. In this work, we propose a neuron group-based structural learning method for an SNN to better capture the spatial and temporal information from the external environment, and propose a time-slicing scheme to better interpret the spatial and temporal information of responses generated by an SNN. Results show that our method indeed helps to enhance the environment perception ability of the SNN, and possesses a certain degree of robustness, enhancing the potential to build an interpretable or understandable AI in the future.
- 27 December 2024
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 81 (1) , 1-24
A new hardware architecture of high-performance real-time texture classification system based on FPGA
- 27 December 2024
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 81 (1) , 1-24
A new hardware architecture of high-performance real-time texture classification system based on FPGA
- 2 December 2024
- Published by Elsevier in Expert Systems with Applications
- Vol. 266
A degradation-aware enhancement network with fused features for fundus images
- 2 December 2024
- Published by Elsevier in Expert Systems with Applications
- Vol. 266
A degradation-aware enhancement network with fused features for fundus images
- Open Access
- 27 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-9
A Sparsity-Aware Autonomous Path Planning Accelerator with Algorithm-Architecture Co-Design
- 27 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-9
Open AccessA Sparsity-Aware Autonomous Path Planning Accelerator with Algorithm-Architecture Co-Design
- Open Access
- 2 April 2024
- Published by Association for Computing Machinery (ACM)
- p. 42
Accelerating Autonomous Path Planning on FPGAs with Sparsity-Aware HW/SW Co-Optimizations
- 2 April 2024
- Published by Association for Computing Machinery (ACM)
- p. 42
Open AccessAccelerating Autonomous Path Planning on FPGAs with Sparsity-Aware HW/SW Co-Optimizations
- Open Access
- 1 January 2023
- Published by International Union of Radio Science (URSI)
Automatic Recognition of Type III Solar Radio Bursts Using Machine Learning
- 1 January 2023
- Published by International Union of Radio Science (URSI)
Open AccessAutomatic Recognition of Type III Solar Radio Bursts Using Machine Learning
- Open Access
- 24 December 2022
- Published by MDPI AG in Bioengineering
- Vol. 10 (1) , 28
A Low Memory Requirement MobileNets Accelerator Based on FPGA for Auxiliary Medical Tasks
Convolutional neural networks (CNNs) have been widely applied in the fields of medical tasks because they can achieve high accuracy in many fields using a large number of parameters and operations. However, many applications designed for auxiliary checks or help need to be deployed into portable devices, where the huge number of operations and parameters of a standard CNN can become an obstruction. MobileNet adopts a depthwise separable convolution to replace the standard convolution, which can greatly reduce the number of operations and parameters while maintaining a relatively high accuracy. Such highly structured models are very suitable for FPGA implementation in order to further reduce resource requirements and improve efficiency. Many other implementations focus on performance more than on resource requirements because MobileNets has already reduced both parameters and operations and obtained significant results. However, because many small devices only have limited resources they cannot run MobileNet-like efficient networks in a normal way, and there are still many auxiliary medical applications that require a high-performance network running in real-time to meet the requirements. Hence, we need to figure out a specific accelerator structure to further reduce the memory and other resource requirements while running MobileNet-like efficient networks. In this paper, a MobileNet accelerator is proposed to minimize the on-chip memory capacity and the amount of data that is transferred between on-chip and off-chip memory. We propose two configurable computing modules: Pointwise Convolution Accelerator and Depthwise Convolution Accelerator, to parallelize the network and reduce the memory requirement with a specific dataflow model. At the same time, a new cache usage method is also proposed to further reduce the use of the on-chip memory. We implemented the accelerator on Xilinx XC7Z020, deployed MobileNetV2 on it, and achieved 70.94 FPS with 524.25 KB on-chip memory usage under 150 MHz.
- 24 December 2022
- Published by MDPI AG in Bioengineering
- Vol. 10 (1) , 28
Open AccessA Low Memory Requirement MobileNets Accelerator Based on FPGA for Auxiliary Medical Tasks
Convolutional neural networks (CNNs) have been widely applied in the fields of medical tasks because they can achieve high accuracy in many fields using a large number of parameters and operations. However, many applications designed for auxiliary checks or help need to be deployed into portable devices, where the huge number of operations and parameters of a standard CNN can become an obstruction. MobileNet adopts a depthwise separable convolution to replace the standard convolution, which can greatly reduce the number of operations and parameters while maintaining a relatively high accuracy. Such highly structured models are very suitable for FPGA implementation in order to further reduce resource requirements and improve efficiency. Many other implementations focus on performance more than on resource requirements because MobileNets has already reduced both parameters and operations and obtained significant results. However, because many small devices only have limited resources they cannot run MobileNet-like efficient networks in a normal way, and there are still many auxiliary medical applications that require a high-performance network running in real-time to meet the requirements. Hence, we need to figure out a specific accelerator structure to further reduce the memory and other resource requirements while running MobileNet-like efficient networks. In this paper, a MobileNet accelerator is proposed to minimize the on-chip memory capacity and the amount of data that is transferred between on-chip and off-chip memory. We propose two configurable computing modules: Pointwise Convolution Accelerator and Depthwise Convolution Accelerator, to parallelize the network and reduce the memory requirement with a specific dataflow model. At the same time, a new cache usage method is also proposed to further reduce the use of the on-chip memory. We implemented the accelerator on Xilinx XC7Z020, deployed MobileNetV2 on it, and achieved 70.94 FPS with 524.25 KB on-chip memory usage under 150 MHz.