Xiaotian Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Xiaotian Wang has a total of 14 co-authors. Here are the top co-authors by number of shared publications:
Letian Zhao
University of Science and Technology of China
4 shared publications- Xi Jin
Department of Physics, University of Science and Technology of China, Hefei, China
4 shared publications - Teng Tian
Hefei University of Technology
3 shared publications Wei Wu
University of Science and Technology of China
3 shared publicationsFangmin Lu
1 shared publication- Huawen Liang
USTC, Hefei, China
1 shared publication
Latest Publications
- Open Access
- 2 April 2024
- Published by Association for Computing Machinery (ACM)
- p. 187
Efficient Message Passing Architecture for GCN Training on HBM-based FPGAs with Orthogonal Topology On-Chip Networks
Graph Convolutional Networks (GCNs) are state-of-the-art deep learning models for representation learning on graphs. However, the efficient training of GCNs is hampered by constraints in memory capacity and bandwidth, compounded by the irregular data flow that results in communication bottlenecks. To address these challenges, we propose a message-passing architecture that leverages NUMA-based memory access properties and employs a parallel multicast routing algorithm based on a 4-D hypercube network within the accelerator for efficient message passing in graphs. Additionally, we have re-engineered the backpropagation algorithm specific to GCNs within our proposed accelerator. This redesign strategically mitigates the memory demands prevalent during the training phase and diminishes the computational overhead associated with the transposition of extensive matrices. Compared to the state-of-the-art HP-GNN architecture we achieved a performance improvement of 1.03×~1.81×.
- 2 April 2024
- Published by Association for Computing Machinery (ACM)
- p. 187
Open AccessEfficient Message Passing Architecture for GCN Training on HBM-based FPGAs with Orthogonal Topology On-Chip Networks
Graph Convolutional Networks (GCNs) are state-of-the-art deep learning models for representation learning on graphs. However, the efficient training of GCNs is hampered by constraints in memory capacity and bandwidth, compounded by the irregular data flow that results in communication bottlenecks. To address these challenges, we propose a message-passing architecture that leverages NUMA-based memory access properties and employs a parallel multicast routing algorithm based on a 4-D hypercube network within the accelerator for efficient message passing in graphs. Additionally, we have re-engineered the backpropagation algorithm specific to GCNs within our proposed accelerator. This redesign strategically mitigates the memory demands prevalent during the training phase and diminishes the computational overhead associated with the transposition of extensive matrices. Compared to the state-of-the-art HP-GNN architecture we achieved a performance improvement of 1.03×~1.81×.
- 29 March 2023
- Published by Springer Nature
- p. 41-53
MCANet: Multiscale Cross-Modality Attention Network for Multispectral Pedestrian Detection
- 29 March 2023
- Published by Springer Nature
- p. 41-53
MCANet: Multiscale Cross-Modality Attention Network for Multispectral Pedestrian Detection
- 15 June 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems II: Express Briefs
- Vol. 69 (11) , 4573-4577
Exploration of Balanced Design in Resource-Constrained Edge Device for Efficient CNNs
Convolutional Neural Networks (CNNs) have been widely used in various image recognition applications due to their high precision and high versatility. In order to improve real-time performance, many efficient CNNs with low computational complexity have been proposed. However, compared with the reduction in the amount of calculation, the demand for memory access has not decreased simultaneously. How to balance the utilization of various resources has become a challenge. This article deeply deconstructs the CNN network and analyzes the relationship among computational resources, on-chip memory, off-chip bandwidth, and throughput. With a given bandwidth, we propose a cross-layer scheduling to minimize the on-chip memory occupancy, which can still fully utilize the computational resources. We implement a flexible FPGA-based hardware accelerator that can deploy the proposed design. According to the experiments on the MobileNetV2, our design reduces the on-chip memory usage by 91.7% and achieves state-of-the-art performance.
- 15 June 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems II: Express Briefs
- Vol. 69 (11) , 4573-4577
Exploration of Balanced Design in Resource-Constrained Edge Device for Efficient CNNs
Convolutional Neural Networks (CNNs) have been widely used in various image recognition applications due to their high precision and high versatility. In order to improve real-time performance, many efficient CNNs with low computational complexity have been proposed. However, compared with the reduction in the amount of calculation, the demand for memory access has not decreased simultaneously. How to balance the utilization of various resources has become a challenge. This article deeply deconstructs the CNN network and analyzes the relationship among computational resources, on-chip memory, off-chip bandwidth, and throughput. With a given bandwidth, we propose a cross-layer scheduling to minimize the on-chip memory occupancy, which can still fully utilize the computational resources. We implement a flexible FPGA-based hardware accelerator that can deploy the proposed design. According to the experiments on the MobileNetV2, our design reduces the on-chip memory usage by 91.7% and achieves state-of-the-art performance.
- 7 June 2022
- Published by Elsevier in Journal of Systems Architecture
- Vol. 129
G-NMP: Accelerating Graph Neural Networks with DIMM-based Near-Memory Processing
- 7 June 2022
- Published by Elsevier in Journal of Systems Architecture
- Vol. 129
G-NMP: Accelerating Graph Neural Networks with DIMM-based Near-Memory Processing
- Open Access
- 1 January 2022
- Published by Elsevier in SSRN Electronic Journal
Fp-Gnn: Adaptive Fpga Accelerator for Graph Neural Networks
Graph Neural Network (GNN) exhibits great success in graph data analysis and promotes the evolution of artificial intelligence. The computing procedure of GNNs can be categorized into two phases: Aggregation and Combination, which have irregular and regular computing characteristics respectively. In this work, we present an adaptive FPGA accelerator, FP-GNN, to enable flexible and efficient GNN acceleration. First, this paper presents quantitative analysis of the impact on performance when changing the order of Aggregation and Combination. Then we propose an Adaptive GNN Accelerator framework (AGA) that adopts a unified processing module to support these two phases simultaneously, and an Adaptive Graph Partition strategy (AGP) to alleviate memory bottleneck and eliminate graph repartition overhead between GNN layers. In addition, we conduct multiple workflow optimizations for both Aggregation and Combination phases to achieve workload balance and feature sparsity elimination. Finally, we implement the FP-GNN on a Xilinx VCU128 FPGA, and conduct experiments on 24 different GNN configurations. Overall, FP-GNN speeds up end-to-end performance by 665x, 24.9x, and 2.33x compared to state-of-the-art CPU, GPU, and FPGA implementations, respectively.
- 1 January 2022
- Published by Elsevier in SSRN Electronic Journal
Open AccessFp-Gnn: Adaptive Fpga Accelerator for Graph Neural Networks
Graph Neural Network (GNN) exhibits great success in graph data analysis and promotes the evolution of artificial intelligence. The computing procedure of GNNs can be categorized into two phases: Aggregation and Combination, which have irregular and regular computing characteristics respectively. In this work, we present an adaptive FPGA accelerator, FP-GNN, to enable flexible and efficient GNN acceleration. First, this paper presents quantitative analysis of the impact on performance when changing the order of Aggregation and Combination. Then we propose an Adaptive GNN Accelerator framework (AGA) that adopts a unified processing module to support these two phases simultaneously, and an Adaptive Graph Partition strategy (AGP) to alleviate memory bottleneck and eliminate graph repartition overhead between GNN layers. In addition, we conduct multiple workflow optimizations for both Aggregation and Combination phases to achieve workload balance and feature sparsity elimination. Finally, we implement the FP-GNN on a Xilinx VCU128 FPGA, and conduct experiments on 24 different GNN configurations. Overall, FP-GNN speeds up end-to-end performance by 665x, 24.9x, and 2.33x compared to state-of-the-art CPU, GPU, and FPGA implementations, respectively.