Lacan: Making Accountability in Anonymous Networks Real
该研究提出Lacan协议，通过创新的验证方法在匿名网络中实现责任追溯，解决了匿名与责任之间的矛盾。
0 citationsRead paper
Selected work
Representative Papers
Accelerating HDC-CNN Hybrid Models Using Custom Instructions on RISC-V GPUs
To address the low energy efficiency and poor programmability of HDC-CNN hybrid models on general-purpose architectures, as well as the lack of native Hyperdimensional Computing (HDC) support in RISC-V GPUs, this paper proposes a customized instruction-set extension for RISC-V GPUs. The extension comprises four instruction categories: HDC encoding, similarity computation, vector operations, and memory access. It enables the first efficient and programmable execution of HDC-CNN hybrid workloads on an open-source RISC-V GPU, balancing flexibility and performance. Microbenchmark results demonstrate up to 56.2× speedup for the extended instructions. End-to-end inference achieves significantly higher energy efficiency and throughput compared to conventional approaches. This work overcomes the flexibility bottleneck of domain-specific accelerators and establishes a new paradigm for high-energy-efficiency hybrid AI computing at the edge.
0 citationsRead paper
Optimizing FPGA and Wafer Test Coverage with Spatial Sampling and Machine Learning
To address the high testing costs for wafers and FPGAs in semiconductor manufacturing and the insufficient prediction robustness of conventional sampling strategies, this paper proposes a spatially optimized machine learning prediction framework. The core innovation is the first-ever Short-Distance Elimination (SDE) algorithm, integrated into two hybrid sampling strategies: S-SDE (combined with stratified sampling) and K-SDE (combined with k-means clustering). By eliminating nearby candidate points, SDE enhances spatial uniformity of training samples, thereby significantly reducing test counts while improving model generalization. Experimental validation on real industrial datasets demonstrates that K-SDE achieves 16.26% (wafer) and 13.07% (FPGA) higher prediction accuracy than standard k-means sampling; S-SDE yields 16.49% (wafer) and 8.84% (FPGA) improvements over stratified sampling. These results confirm the framework’s superior accuracy, robustness, and engineering practicality.
0 citationsRead paper
Vertex evaluation of multiplex graphs using Forman Curvature
Assessing vertex centrality in multilayer networks remains challenging due to the complexity of inter-layer dependencies and heterogeneous topologies. Method: This paper introduces the first extension of Forman curvature to multiplex graphs, proposing a layer-aware curvature definition grounded in discrete differential geometry and multiplex graph modeling. We develop an efficient computational framework that captures intrinsic relationships between curvature, vertex centrality, and global structural properties. Contribution/Results: The proposed curvature metric enables simultaneous vertex importance ranking and network-type classification. Extensive experiments on real-world multilayer networks demonstrate that it significantly outperforms conventional centrality measures—achieving higher accuracy in critical node identification and structural discrimination. This work establishes a novel geometric paradigm for analyzing complex multilayer systems and provides a scalable, theoretically principled tool for network science.
0 citationsRead paper
Recent publications
Latest Papers
Lacan: Making Accountability in Anonymous Networks Real
该研究提出Lacan协议，通过创新的验证方法在匿名网络中实现责任追溯，解决了匿名与责任之间的矛盾。
0 citationsRead paper
Accelerating HDC-CNN Hybrid Models Using Custom Instructions on RISC-V GPUs
To address the low energy efficiency and poor programmability of HDC-CNN hybrid models on general-purpose architectures, as well as the lack of native Hyperdimensional Computing (HDC) support in RISC-V GPUs, this paper proposes a customized instruction-set extension for RISC-V GPUs. The extension comprises four instruction categories: HDC encoding, similarity computation, vector operations, and memory access. It enables the first efficient and programmable execution of HDC-CNN hybrid workloads on an open-source RISC-V GPU, balancing flexibility and performance. Microbenchmark results demonstrate up to 56.2× speedup for the extended instructions. End-to-end inference achieves significantly higher energy efficiency and throughput compared to conventional approaches. This work overcomes the flexibility bottleneck of domain-specific accelerators and establishes a new paradigm for high-energy-efficiency hybrid AI computing at the edge.
0 citationsRead paper
Optimizing FPGA and Wafer Test Coverage with Spatial Sampling and Machine Learning
To address the high testing costs for wafers and FPGAs in semiconductor manufacturing and the insufficient prediction robustness of conventional sampling strategies, this paper proposes a spatially optimized machine learning prediction framework. The core innovation is the first-ever Short-Distance Elimination (SDE) algorithm, integrated into two hybrid sampling strategies: S-SDE (combined with stratified sampling) and K-SDE (combined with k-means clustering). By eliminating nearby candidate points, SDE enhances spatial uniformity of training samples, thereby significantly reducing test counts while improving model generalization. Experimental validation on real industrial datasets demonstrates that K-SDE achieves 16.26% (wafer) and 13.07% (FPGA) higher prediction accuracy than standard k-means sampling; S-SDE yields 16.49% (wafer) and 8.84% (FPGA) improvements over stratified sampling. These results confirm the framework’s superior accuracy, robustness, and engineering practicality.
0 citationsRead paper
Vertex evaluation of multiplex graphs using Forman Curvature
Assessing vertex centrality in multilayer networks remains challenging due to the complexity of inter-layer dependencies and heterogeneous topologies. Method: This paper introduces the first extension of Forman curvature to multiplex graphs, proposing a layer-aware curvature definition grounded in discrete differential geometry and multiplex graph modeling. We develop an efficient computational framework that captures intrinsic relationships between curvature, vertex centrality, and global structural properties. Contribution/Results: The proposed curvature metric enables simultaneous vertex importance ranking and network-type classification. Extensive experiments on real-world multilayer networks demonstrate that it significantly outperforms conventional centrality measures—achieving higher accuracy in critical node identification and structural discrimination. This work establishes a novel geometric paradigm for analyzing complex multilayer systems and provides a scalable, theoretically principled tool for network science.
0 citationsRead paper