Li Zeng
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
Li Zeng has a total of 10 co-authors. Here are the top co-authors by number of shared publications:
Christian Jansen
Univ. Hospital Basel (Switzerland)
1 shared publication- Haoran Cai
AI Data Department, GTS, Huawei Technologies Co., Ltd
1 shared publication - Jinhua Zhou
Huawei Technologies Co Ltd
1 shared publication - Lei Zou
Peking University, Beijing, China
1 shared publication - M. Tamer Özsu
University of Waterloo Canada
1 shared publication - Michael Unser
Biomedical Imaging Group, École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland
1 shared publication
Latest Publications
- 20 December 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Knowledge and Data Engineering
- Vol. 35 (11) , 11899-11916
SGSI – A Scalable GPU-Friendly Subgraph Isomorphism Algorithm
Due to the inherent hardness of subgraph isomorphism, the performance is often a bottleneck in various real-world applications. We address this by designing an efficient subgraph isomorphism algorithm leveraging features of GPU architecture. Existing GPU-based solutions adopt two-step output scheme, performing the same join twice in order to write intermediate results concurrently. They also lack GPU architecture-aware optimizations that allow scaling to large graphs. In this paper, we propose a S calable G PU-friendly s ubgraph i somorphism algorithm, SGSI . SGSI incorporates a Prealloc-Combine strategy based on the vertex-oriented framework, which avoids joining-twice in existing solutions. It uses a GPU-friendly data structure (called PCSR ) to represent an edge-labeled graph. We also study fine-grained load balance strategies and discuss how to handle enormous graphs that cannot be resident in GPU memory. A partition-based pipeline framework is proposed. Extensive experiments on both synthetic and real graphs show that SGSI outperforms the state-of-the-art algorithms by up to several orders of magnitude and has a good scalability with graph size scaling to billions of edges.
- 20 December 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Knowledge and Data Engineering
- Vol. 35 (11) , 11899-11916
SGSI – A Scalable GPU-Friendly Subgraph Isomorphism Algorithm
Due to the inherent hardness of subgraph isomorphism, the performance is often a bottleneck in various real-world applications. We address this by designing an efficient subgraph isomorphism algorithm leveraging features of GPU architecture. Existing GPU-based solutions adopt two-step output scheme, performing the same join twice in order to write intermediate results concurrently. They also lack GPU architecture-aware optimizations that allow scaling to large graphs. In this paper, we propose a S calable G PU-friendly s ubgraph i somorphism algorithm, SGSI . SGSI incorporates a Prealloc-Combine strategy based on the vertex-oriented framework, which avoids joining-twice in existing solutions. It uses a GPU-friendly data structure (called PCSR ) to represent an edge-labeled graph. We also study fine-grained load balance strategies and discuss how to handle enormous graphs that cannot be resident in GPU memory. A partition-based pipeline framework is proposed. Extensive experiments on both synthetic and real graphs show that SGSI outperforms the state-of-the-art algorithms by up to several orders of magnitude and has a good scalability with graph size scaling to billions of edges.
- Open Access
- 5 December 2001
- Published by SPIE-Intl Soc Optical Eng
- Vol. 4478, 427-433
Extension of wavelet compression algorithms to 3D and 4D image data: exploitation of data coherence in higher dimensions allows very high compression ratios
High resolution multidimensional image data yield huge datasets. For compression and analysis, 2D approaches are often used, neglecting the information coherence in higher dimensions, which can be exploited for improved compression. We designed a wavelet compression algorithm suited for data of arbitrary dimensions, and assessed its ability for compression of 4D medical images. Basically, separable wavelet transforms are done in each dimension, followed by quantization and standard coding. Results were compared with conventional 2D wavelet. We found that in 4D heart images, this algorithm allowed high compression ratios, preserving diagnostically important image features. For similar image quality, compression ratios using the 3D/4D approaches were typically much higher (2-4 times per added dimension) than with the 2D approach. For low-resolution images created with the requirement to keep predefined key diagnostic information (contractile function of the heart), compression ratios up to 2000 could be achieved. Thus, higher-dimensional wavelet compression is feasible, and by exploitation of data coherence in higher image dimensions allows much higher compression than comparable 2D approaches. The proven applicability of this approach to multidimensional medical imaging has important implications especially for the fields of image storage and transmission and, specifically, for the emerging field of telemedicine.
- 5 December 2001
- Published by SPIE-Intl Soc Optical Eng
- Vol. 4478, 427-433
Open AccessExtension of wavelet compression algorithms to 3D and 4D image data: exploitation of data coherence in higher dimensions allows very high compression ratios
High resolution multidimensional image data yield huge datasets. For compression and analysis, 2D approaches are often used, neglecting the information coherence in higher dimensions, which can be exploited for improved compression. We designed a wavelet compression algorithm suited for data of arbitrary dimensions, and assessed its ability for compression of 4D medical images. Basically, separable wavelet transforms are done in each dimension, followed by quantization and standard coding. Results were compared with conventional 2D wavelet. We found that in 4D heart images, this algorithm allowed high compression ratios, preserving diagnostically important image features. For similar image quality, compression ratios using the 3D/4D approaches were typically much higher (2-4 times per added dimension) than with the 2D approach. For low-resolution images created with the requirement to keep predefined key diagnostic information (contractile function of the heart), compression ratios up to 2000 could be achieved. Thus, higher-dimensional wavelet compression is feasible, and by exploitation of data coherence in higher image dimensions allows much higher compression than comparable 2D approaches. The proven applicability of this approach to multidimensional medical imaging has important implications especially for the fields of image storage and transmission and, specifically, for the emerging field of telemedicine.