Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

Accelerating Sparse Graph Neural Networks with Tensor Core Optimization

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read FTC-GNN accelerates sparse GNNs by running Tensor Cores and CUDA Cores in parallel, reporting kernel-time speedups of up to 7.10x over established libraries.

desk verdict The abstract's 5.32x AGNN-vs-DGL speedup is contradicted by the paper's own Table VI, and the novelty over TC-GNN is thin, but the GCN data and implementation description give the paper enough substance to justify a corrective review round. read the letter →

arxiv 2412.12218 v2 pith:FWDVLCKH submitted 2024-12-16 cs.LG cs.AR

Ka Wai Wu This is my paper · ORCID
classification cs.LGcs.AR
keywords graphneuralnetworksTensorCoresCUDAsparsematrixmultiplicationkernelfusionGPUaccelerationGCNAGNN
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

FTC-GNN is a framework for accelerating sparse graph neural networks by running Tensor Cores and CUDA Cores simultaneously instead of one after the other. Its sparse-to-dense transformation packs the irregular adjacency matrix into dense blocks that Tensor Cores multiply efficiently, while CUDA Cores handle data management and the remaining sparse edges. Using GCN and AGNN models on five graph datasets, the paper reports average GPU kernel-time speedups of 4.90x and 7.10x for GCN against DGL and PyG, 1.17x against TC-GNN, and corresponding AGNN speedups of 5.32x, 2.92x, and 1.02x. The result matters because sparse graph workloads typically leave GPUs underutilized, and the design shows a practical way to keep both compute units busy.

What carries the argument

The load-bearing mechanism is the sparse graph transformation technique, which reads the graph in CSR form, groups rows into row windows, compresses duplicate neighbor loads, and emits dense TC blocks with block height 16 and width 8. These blocks are consumed by a single fused CUDA kernel that uses WMMA instructions to perform GEMM on Tensor Cores for most blocks and scalar multiply-add on CUDA Cores for the remainder, with shared memory acting as the buffer that both units read from. The same transformation feeds two kernels: sparse neighbor aggregation, an SpMM-style operation, and sparse edge feature computation, an SDDMM-style operation.

What would settle it

Rerun the FTC-GNN, DGL, PyG, and TC-GNN comparison on the same five datasets but measure full end-to-end time including the sparse graph transformation and, for training, the accuracy after convergence. If the transformation cost is included, small graphs such as citeseer and cora should show much smaller speedups; recomputing the AGNN-versus-DGL average from Table VI will also reveal whether the abstract's 5.32x figure can be reproduced.

Watch

Extended reading notes

Core claim

The paper's central claim is that the usual bottleneck in sparse GNN acceleration is not the raw speed of either unit but the failure to use them concurrently. FTC-GNN therefore divides the work by data shape: dense matrix-multiply fragments are sent to Tensor Cores, while CUDA Cores execute the memory-intensive and irregular parts of the computation. The measured evidence is a set of GPU kernel-time comparisons, presented per dataset, showing that FTC-GNN is faster than DGL and PyG on all five datasets under GCN, and faster than PyG under AGNN, while remaining competitive with TC-GNN. The authors state the speedups as 4.90x, 7.10x, and 1.17x for GCN and 5.32x, 2.92x, and 1.02x for AGNN relative to DGL, PyG, and TC-GNN.

Load-bearing premise

The load-bearing premise is that average GPU kernel time alone is a fair and complete measure of acceleration, so the one-time sparse graph transformation cost, end-to-end runtime, and model accuracy can be left out of the speedup comparison.

Editorial extensions

If this is right

  • A one-time graph transformation can be reused across many training and inference iterations, so the preprocessing cost is amortised rather than paid per layer.
  • The same CUDA-plus-Tensor-Core split can accelerate both SpMM-like neighbor aggregation and SDDMM-like edge feature computation, covering the two main sparse operations in GNN layers.
  • Because the split is done inside one kernel, FTC-GNN avoids the overhead of launching separate kernels for the dense and sparse portions of the computation.
  • If the kernel-time speedups hold, GNN frameworks can adopt the technique without changing model semantics, only the internal scheduling of matrix work.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A reader redoing the arithmetic on Table VI will find that the AGNN-versus-DGL average is not the 5.32x stated in the abstract.
  • A testable extension would be to measure end-to-end time including the one-time sparse transformation; small graphs such as citeseer and cora would likely show smaller speedups than the kernel-only figures.
  • The same heterogeneous scheduling idea could generalize to other sparse deep learning workloads, such as sparse attention or recommendation models that mix irregular indexing with dense matrix multiplication.
  • An adaptive version that sizes the CUDA/Tensor-Core split per row window or per graph density is a natural next step for graphs where TC blocks are few.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes FTC-GNN, a GPU framework for sparse graph neural network computation that aims to use Tensor Cores and CUDA Cores in parallel. It introduces a sparse-to-dense graph transformation technique, sparse neighbor aggregation, and sparse edge feature computation kernels built on the WMMA API, and reports speedups over DGL, PyG, and TC-GNN for GCN and AGNN models on five datasets. The abstract claims average speedups of 4.90x, 7.10x, and 1.17x for GCN and 5.32x, 2.92x, and 1.02x for AGNN against DGL, PyG, and TC-GNN, respectively.

Significance. If the reported speedups were reproducible, FTC-GNN would be a practically useful contribution: it targets a real bottleneck, the poor utilization of Tensor Cores for irregular sparse GNN workloads, and it goes beyond existing work such as TC-GNN by attempting a more balanced partitioning between Tensor Cores and CUDA Cores. The paper deserves credit for reporting per-dataset kernel-time tables, which allow an independent reader to recompute the claimed averages, and for comparing against multiple established baselines. However, the central performance claim for half of the evaluation is directly contradicted by the paper's own data: the abstract's 5.32x AGNN-vs-DGL speedup is not reproducible from Table VI, which shows FTC-GNN slower than DGL on amazon0505 and a per-dataset average speedup of about 1.97x. In addition, the evaluation is limited to GPU kernel time, excluding the one-time sparse graph transformation cost, and no accuracy results are provided despite the stated requirement to preserve accuracy. These gaps mean the practical superiority of FTC-GNN is not currently established.

major comments (3)
  1. [Abstract, Section VI.B.1, Table VI, Figure 11] The abstract states that for AGNN, FTC-GNN achieves a 5.32x speedup over DGL, and Section VI.B.1 repeats this claim. This is not supported by the data in Table VI. Computing the ratio DGL average GPU kernel time divided by FTC-GNN average GPU kernel time for each dataset gives citeseer 5.673/1.665 = 3.41x, cora 5.630/1.744 = 3.23x, amazon0505 35.841/38.762 = 0.93x, com-amazon 21.269/19.179 = 1.11x, and amazon0601 26.754/22.358 = 1.20x. The arithmetic mean of these per-dataset ratios is about 1.97x, and the ratio of summed DGL time to summed FTC-GNN time is about 1.14x; neither equals 5.32x. The narrative in Section VI.B.1 is also inconsistent with its own table: it says FTC-GNN's average GPU kernel time is 'significantly higher' than DGL for amazon0505, com-amazon, and amazon0601, but Table VI shows FTC-GNN is faster than DGL on com-amazon (19.179 ms vs. 21.269 ms) and amazon0601 (22.358 ms vs. 26.754 ms), and slower only on amazon0505. This is a direct internal contradiction that invalidates the headline AGNN-vs-DGL result.
  2. [Section VI.B.1 and Section V.A.1] All speedup claims are based on average GPU kernel time only, as defined in Section VI.B.1, and exclude the one-time sparse graph transformation cost described in Section V.A.1. The transformation is a required preprocessing step that converts the input CSR graph into TC blocks, so a practical performance comparison should include it or provide an amortized analysis. For small datasets such as citeseer and cora, the reported FTC-GNN kernel times are sub-millisecond (0.551 ms and 0.485 ms for GCN), so even a modest transformation overhead could dominate the kernel time and change the direction of the comparison. No end-to-end runtime or total training time is reported anywhere in Section VI. The claimed practical acceleration therefore rests on the unstated assumption that kernel time alone is a fair and complete proxy for GNN performance.
  3. [Section III.C and Section VI] Section III.C states that the acceleration approach 'must ensure' that the accuracy of GNN models is not compromised, but the experimental section contains no accuracy measurements for any model or dataset. This is a load-bearing omission because the implementation uses TF-32 and half-precision WMMA operations (Sections V.A.1 and V.A.2), which can change numerical results compared to FP32. Without accuracy results, the reader cannot verify that the reported speedups are achieved at acceptable model quality, and the stated research goal of preserving accuracy is left unvalidated.
minor comments (5)
  1. [Section VI.B.1, Table VI] The sentence 'FTC-GNN and DGL show relatively similar performance in terms of average GPU kernel time' is misleading: per Table VI, FTC-GNN is about 3.4x and 3.2x faster than DGL on citeseer and cora, respectively, for AGNN.
  2. [Figures 10-15] The figures are referenced in the text but not reproduced in the manuscript, and the axis definitions are not described. In particular, Figure 11 is claimed to show a 5.32x average speedup that does not agree with Table VI; if the figure aggregates data differently, the aggregation rule should be stated explicitly.
  3. [Section V.A.1 and V.A.2] The precision used in the experiments is ambiguous: Section V.A.1 says the TF-32 data format is adopted with TC_BLK_H = 16 and TC_BLK_W = 8, while Section V.A.2 mentions that half-precision floating-point numbers are used to reduce complexity. The relationship between these two statements and the actual precision used in the reported kernel times should be clarified.
  4. [Table III] The hardware table lists a 'Graphics Card Cirrus Logic GD 5446' alongside the Tesla V100; this appears to be a display adapter rather than the device used for CUDA computation, and the entry should be removed or clarified.
  5. [References] Several references appear unrelated to the claims they support, for example the TechRxiv preprints on housing price prediction and Wikipedia topic modeling cited in the context of recommendation systems and social networks; the authors should verify that each citation matches its context.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular reasoning: FTC-GNN's claims are empirical benchmark comparisons against external systems, not derived from fitted inputs or self-citations.

full rationale

The paper contains no derivation chain in which an output is defined in terms of the claimed result. FTC-GNN's central claims are speedup measurements against DGL, PyG, and TC-GNN, computed as ratios of measured GPU kernel times. These baselines are external systems; the measurements are reported in tables and are not produced by fitting a parameter to the target result. The sparse-to-dense transformation and CUDA/Tensor Core kernel design are engineering contributions evaluated empirically, so no 'prediction' is forced by construction. The author's self-citations concern unrelated topics and are not load-bearing for the acceleration claims. The abstract's AGNN-vs-DGL speedup of 5.32x is not reproduced by Table VI, and the evaluation excludes one-time transformation overhead and accuracy; these are serious correctness concerns, but not circularity. Accordingly the circularity score is 0.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central support for the paper's performance claim is the benchmark protocol. That protocol rests on assumptions about what time is counted, what precision is acceptable, and which hardware and software stack is used. These assumptions are listed above.

free parameters (2)
  • TC_BLK_H and TC_BLK_W = 16 and 8
    Kernel tile dimensions set in Section V.A.1; they determine WMMA fragment shapes and affect all measured kernel timings, but no sensitivity analysis or tuning procedure is reported.
  • TCU/CUDA work partition = not specified
    Section V.A.2 splits TC blocks between Tensor Cores and CUDA Cores, describing the split as 'majority' vs 'remainder'; the actual ratio is a hand-chosen implementation choice that affects the speedups but is never quantified.
assumptions (4)
  • domain assumption Average GPU kernel time, excluding the one-time sparse graph transformation, is the correct performance metric for comparing GNN frameworks.
    Speedup ratios in Section VI.B.1 are computed from GPU kernel time only; transformation and preprocessing costs are not included, which can bias results for small graphs like citeseer and cora.
  • domain assumption Tensor Core mixed-precision computation preserves GNN accuracy.
    Section III.C states that accuracy must not be compromised, but Section VI reports no accuracy numbers or precision comparison.
  • domain assumption The experimental GPU behaves as a standard Tesla V100 with CUDA 11.3 and WMMA support.
    Used in Section VI.A; performance results are hardware-specific and may not transfer to other GPU architectures.
  • domain assumption The five benchmark datasets and two GNN models are representative of sparse GNN workloads.
    Datasets in Table IV are the same families used by TC-GNN; no analysis of degree distribution, feature dimensions, or workload diversity is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Accelerating Sparse Graph Neural Networks with Tensor Core Optimization." pith.science (2026). https://pith.science/paper/FWDVLCKH

@misc{pith2026241212218,
  author       = {Pith},
  title        = {Pith review of: Accelerating Sparse Graph Neural Networks with Tensor Core Optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FWDVLCKH}},
  note         = {Machine review of arXiv:2412.12218}
}
read the original abstract

Graph neural networks (GNNs) have seen extensive application in domains such as social networks, bioinformatics, and recommendation systems. However, the irregularity and sparsity of graph data challenge traditional computing methods, which are insufficient to meet the performance demands of GNNs. Recent research has explored parallel acceleration using CUDA Cores and Tensor Cores, but significant challenges persist: (1) kernel fusion leads to false high utilization, failing to treat CUDA and Tensor Cores as independent resources, and (2) heterogeneous cores have distinct computation preferences, causing inefficiencies. To address these issues, this paper proposes FTC-GNN, a novel acceleration framework that efficiently utilizes CUDA and Tensor Cores for GNN computation. FTC-GNN introduces (1) a collaborative design that enables the parallel utilization of CUDA and Tensor Cores and (2) a sparse-to-dense transformation strategy that assigns dense matrix operations to Tensor Cores while leveraging CUDA Cores for data management and sparse edge processing. This design optimizes GPU resource utilization and improves computational efficiency. Experimental results demonstrate the effectiveness of FTC-GNN using GCN and AGNN models across various datasets. For GCN, FTC-GNN achieves speedups of 4.90x, 7.10x, and 1.17x compared to DGL, PyG, and TC-GNN, respectively. For AGNN, it achieves speedups of 5.32x, 2.92x, and 1.02x, establishing its superiority in accelerating GNN computations.

Figures

Figures reproduced from arXiv: 2412.12218 by the authors.

Figure 1
Figure 1. Computational Process of GNN In the aggregation phase, the main task of a GNN is to aggregate information from each node in the irregular input graph and its neighboring nodes. In the update phase, the GNN utilizes the local representations obtained from the aggregation phase to update the feature representations of each node. The basic formula for the aggregation phase is shown in Equation 1, while the basic formul… view at source ↗
Figure 2
Figure 2. Active Scheduling of Tensor Cores and CUDA Cores [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. TC-GNN also implements a collaborative design between CUDA and Tensor Core Units (TCUs), mixing the CUDA Core’s SIMT work mode with TCU’s warp processing into a single GPU kernel. Threads from the same thread block running on CUDA Cores load data from global memory into shared memory. Once data loading is completed on CUDA Cores, each warp’s threads operate on TCUs for GEMM com￾putation. TC-GNN is fully integrated w… view at source ↗
Figures from the paper (12 more)
Figure 3
Figure 3. Figure 3: Dense Block Representation GraphSAGE samples a fixed number of nodes from a node’s neighbors at each layer, aggregates their feature information, and generates the feature representation for the current node, as illustrated in [PITH_FULL_IMAGE:figures/full_fig_p004_3.png]
Figure 4
Figure 4. Figure 4: GraphSAGE Sampling Illustration A new graph convolutional network architecture, Spectral Graph Attention Network (SGAT), was proposed, incorporat￾ing a Fast Eigen-Approximation algorithm for graph convo￾lution operations. SGAT, based on spectral domain methods, perform…
Figure 5
Figure 5. Figure 5: CSR Representation of a Matrix 2) Low Memory Access Efficiency: Although the hybrid sparse-dense scheme addresses some of the aforementioned issues, it still suffers from high overhead and computational waste. First, the distribution of non-zero elements in the adjacen…
Figure 6
Figure 6. Figure 6: Flowchart of Sparse Graph Transformation Technique [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Example of Sparse Graph Transformation Technique [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 8
Figure 8. Figure 8: Flowchart of Sparse Neighbor Aggregation [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 9
Figure 9. Figure 9: Flowchart of Edge Feature Computation used for building deep learning models. In this study, PyTorch was used to write code for accelerating sparse GNNs, as well as for training and testing the model using the tools and functions provided by PyTorch. PyTorch offers a w…
Figure 10
Figure 10. Figure 10: Comparison of Average GPU Kernel Time Between FTC-GNN and [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]
Figure 11
Figure 11. Figure 11: Comparison of Average GPU Kernel Time Between FTC-GNN and [PITH_FULL_IMAGE:figures/full_fig_p013_11.png]
Figure 13
Figure 13. Figure 13: Comparison of Average GPU Kernel Time Between FTC-GNN and [PITH_FULL_IMAGE:figures/full_fig_p014_13.png]
Figure 15
Figure 15. Figure 15: Comparison of Average GPU Kernel Time Between FTC-GNN and [PITH_FULL_IMAGE:figures/full_fig_p014_15.png]
Figure 14
Figure 14. Figure 14: Comparison of Average GPU Kernel Time Between FTC-GNN and [PITH_FULL_IMAGE:figures/full_fig_p014_14.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Blockchain-Based Secure Vehicle Auction System with Smart Contracts

    cs.CR 2025-01 reject novelty 2.0 of 10

    A student prototype applies Ethereum smart contracts to used-car auctions, claiming security and privacy gains, but ships no code and offers no comparison to existing systems.

Reference graph

Works this paper leans on

41 extracted references · 31 canonical work pages · cited by 1 Pith paper

  1. [1]

    A survey of graph neural networks for social recommender systems,

    K. Sharma, Y .-C. Lee, S. Nambi, A. Salian, S. Shah, S.-W. Kim, and S. Kumar, “A survey of graph neural networks for social recommender systems,” ACM Computing Surveys , vol. 56, no. 10, pp. 1–34, 2024, ACM, New York, NY

    work page 2024

  2. [2]

    Graph repre- sentation learning in bioinformatics: trends, methods and applications,

    H.-C. Yi, Z.-H. You, D.-S. Huang, and C. K. Kwoh, “Graph repre- sentation learning in bioinformatics: trends, methods and applications,” Briefings in Bioinformatics , vol. 23, no. 1, Article bbab340, 2022, Oxford University Press

    work page 2022

  3. [3]

    Finding patient zero: Learning contagion source with graph neural networks,

    C. Shah, N. Dehmamy, N. Perra, M. Chinazzi, A.-L. Barab ´asi, A. Vespignani, and R. Yu, “Finding patient zero: Learning contagion source with graph neural networks,” arXiv preprint arXiv:2006.11913 , 2020

    arXiv 2006

  4. [4]

    Graph convolutional networks for computational drug development and discovery,

    M. Sun, S. Zhao, C. Gilvary, O. Elemento, J. Zhou, and F. Wang, “Graph convolutional networks for computational drug development and discovery,” Briefings in Bioinformatics , vol. 21, no. 3, pp. 919–935, 2020, Oxford University Press

    work page 2020

  5. [5]

    Heterogeneous graph neural networks for malicious account detection,

    Z. Liu, C. Chen, X. Yang, J. Zhou, X. Li, and L. Song, “Heterogeneous graph neural networks for malicious account detection,” in Proceedings of the 27th ACM International Conference on Information and Knowl- edge Management, 2018, pp. 2077–2085

    work page 2018

  6. [6]

    Graph neural networks for social recommendation,

    W. Fan, Y . Ma, Q. Li, Y . He, E. Zhao, J. Tang, and D. Yin, “Graph neural networks for social recommendation,” in Proceedings of The World Wide Web Conference, 2019, pp. 417–426

    work page 2019

  7. [7]

    Graph convolutional neural networks for web-scale rec- ommender systems,

    R. Ying, R. He, K. Chen, P. Eksombatchai, W. L. Hamilton, and J. Leskovec, “Graph convolutional neural networks for web-scale rec- ommender systems,” in Proceedings of the 24th ACM SIGKDD Interna- tional Conference on Knowledge Discovery & Data Mining , 2018, pp. 974–983

    work page 2018

  8. [8]

    Hygcn: A GCN accelerator with hybrid architecture,

    M. Yan, L. Deng, X. Hu, L. Liang, Y . Feng, X. Ye, Z. Zhang, D. Fan, and Y . Xie, “Hygcn: A GCN accelerator with hybrid architecture,” in Proceedings of the 2020 IEEE International Symposium on High Performance Computer Architecture (HPCA) , 2020, pp. 15–29, IEEE

    work page 2020

Show all 41 references
  1. [9]

    Strengthening DeFi Security: A Static Analysis Approach to Flash Loan Vulnerabilities,

    K. W. Wu, “Strengthening DeFi Security: A Static Analysis Approach to Flash Loan Vulnerabilities,” *arXiv preprint arXiv:2411.01230*, 2024

    2024 arXiv

  2. [10]

    TC-GNN: Bridging sparse GNN computation and dense tensor cores on GPUs,

    Y . Wang, B. Feng, Z. Wang, G. Huang, and Y . Ding, “TC-GNN: Bridging sparse GNN computation and dense tensor cores on GPUs,” in Proceedings of the 2023 USENIX Annual Technical Conference (USENIX ATC 23) , 2023, pp. 149–164

    2023

  3. [11]

    Graph neural networks in node classification: survey and evaluation,

    S. Xiao, S. Wang, Y . Dai, and W. Guo, “Graph neural networks in node classification: survey and evaluation,” Machine Vision and Applications, vol. 33, no. 1, Article 4, 2022, Springer

    2022

  4. [12]

    Clnode: Curriculum learning for node classification,

    X. Wei, X. Gong, Y . Zhan, B. Du, Y . Luo, and W. Hu, “Clnode: Curriculum learning for node classification,” in Proceedings of the Sixteenth ACM International Conference on Web Search and Data Mining, 2023, pp. 670–678

    2023

  5. [13]

    Hermes: Boosting the Perfor- mance of Machine-Learning-Based Intrusion Detection System through Geometric Feature Learning,

    C. Zhang, S. Shi, N. Wang, X. Xu, S. Li, L. Zheng, R. Marchany, M. Gardner, Y . T. Hou, and W. Lou, “Hermes: Boosting the Perfor- mance of Machine-Learning-Based Intrusion Detection System through Geometric Feature Learning,” in *Proceedings of the Twenty-fifth Inter- national...

    2024

  6. [14]

    Explicit pairwise factorized graph neural network for semi-supervised node classification,

    Y . Wang, Y . Shen, and D. Cremers, “Explicit pairwise factorized graph neural network for semi-supervised node classification,” in Proceedings of Uncertainty in Artificial Intelligence , 2021, pp. 1979–1987, PMLR

    2021

  7. [15]

    Link prediction based on graph neural net- works,

    M. Zhang and Y . Chen, “Link prediction based on graph neural net- works,” Advances in Neural Information Processing Systems , vol. 31, 2018

    2018

  8. [16]

    Advanced Machine Learning for Housing Market Analysis: Predicting Property Prices in Washington, D.C

    Pei-Hsuan Sung. Advanced Machine Learning for Housing Market Analysis: Predicting Property Prices in Washington, D.C. TechRxiv. January 04, 2025. DOI: 10.36227/techrxiv.173603502.26172421/v1

    2025

  9. [17]

    Temporal network embedding for link prediction via V AE joint attention mechanism,

    P. Jiao, X. Guo, X. Jing, D. He, H. Wu, S. Pan, M. Gong, and W. Wang, “Temporal network embedding for link prediction via V AE joint attention mechanism,” IEEE Transactions on Neural Networks and Learning Systems, vol. 33, no. 12, pp. 7400–7413, 2021, IEEE

    2021

  10. [18]

    Hashing-accelerated graph neural networks for link prediction,

    W. Wu, B. Li, C. Luo, and W. Nejdl, “Hashing-accelerated graph neural networks for link prediction,” in Proceedings of the Web Conference 2021, 2021, pp. 2910–2920

    2021

  11. [19]

    Net: Degree-specific graph neural networks for node and graph classification,

    J. Wu, J. He, and J. Xu, “Net: Degree-specific graph neural networks for node and graph classification,” in Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2019, pp. 406–415

    2019

  12. [20]

    Multi-scale graph classification with shared graph neural network,

    P. Zhou, Z. Wu, G. Wen, K. Tang, and J. Ma, “Multi-scale graph classification with shared graph neural network,” World Wide Web, vol. 26, no. 3, pp. 949–966, 2023, Springer

    2023

  13. [21]

    On size-oriented long- tailed graph classification of graph neural networks,

    Z. Liu, Q. Mao, C. Liu, Y . Fang, and J. Sun, “On size-oriented long- tailed graph classification of graph neural networks,” in Proceedings of the ACM Web Conference 2022 , 2022, pp. 1506–1516

    2022

  14. [22]

    Cusparse library,

    M. Naumov, L. Chien, P. Vandermersch, and U. Kapasi, “Cusparse library,” in Proceedings of the GPU Technology Conference , vol. 12, 2010

    2010

  15. [23]

    Algorithm 1000: SuiteSparse: GraphBLAS: Graph algo- rithms in the language of sparse linear algebra,

    T. A. Davis, “Algorithm 1000: SuiteSparse: GraphBLAS: Graph algo- rithms in the language of sparse linear algebra,” ACM Transactions on Mathematical Software (TOMS) , vol. 45, no. 4, pp. 1–25, 2019, ACM, New York, NY , USA

    2019

  16. [24]

    High-performance computing on the Intel Xeon Phi,

    E. Wang, Q. Zhang, B. Shen, G. Zhang, X. Lu, Q. Wu, and Y . Wang, “High-performance computing on the Intel Xeon Phi,” Springer, vol. 5, p. 2, 2014, Springer

    2014

  17. [25]

    Improving GPU throughput through parallel execution using tensor cores and CUDA cores,

    K. Ho, H. Zhao, A. Jog, and S. Mohanty, “Improving GPU throughput through parallel execution using tensor cores and CUDA cores,” in Proceedings of the 2022 IEEE Computer Society Annual Symposium on VLSI (ISVLSI) , 2022, pp. 223–228, IEEE

    2022

  18. [26]

    Community Structure and Connectivity Analysis in So- cial Networks Using Community Detection Methods

    Pei-Hsuan Sung. Community Structure and Connectivity Analysis in So- cial Networks Using Community Detection Methods. TechRxiv. January 21, 2025. DOI: 10.36227/techrxiv.173750116.69554018/v1

    2025

  19. [27]

    Inductive representation learning on large graphs,

    W. Hamilton, Z. Ying, and J. Leskovec, “Inductive representation learning on large graphs,” Advances in Neural Information Processing Systems, vol. 30, 2017

    2017

  20. [28]

    Fast batched matrix multiplication for small sizes using half-precision arithmetic on GPUs,

    A. Abdelfattah, S. Tomov, and J. Dongarra, “Fast batched matrix multiplication for small sizes using half-precision arithmetic on GPUs,” in Proceedings of the 2019 IEEE International Parallel and Distributed Processing Symposium (IPDPS) , 2019, pp. 111–122, IEEE

    2019

  21. [29]

    Tacker: Tensor-CUDA core kernel fusion for improving the GPU utilization while ensuring QoS,

    H. Zhao, W. Cui, Q. Chen, Y . Zhang, Y . Lu, C. Li, J. Leng, and M. Guo, “Tacker: Tensor-CUDA core kernel fusion for improving the GPU utilization while ensuring QoS,” in Proceedings of the 2022 IEEE International Symposium on High-Performance Computer Architecture (HPCA), 202...

    2022

  22. [30]

    Blockchain-Based Secure Vehicle Auction System with Smart Contracts

    Ka Wai Wu. Blockchain-Based Secure Vehicle Auction System with Smart Contracts . arXiv, January 2025. DOI: 10.48550/arXiv.2501.04841

    2025 doi arXiv

  23. [31]

    Ge-spmm: General-purpose sparse matrix-matrix multiplication on GPUs for graph neural networks,

    G. Huang, G. Dai, Y . Wang, and H. Yang, “Ge-spmm: General-purpose sparse matrix-matrix multiplication on GPUs for graph neural networks,” in Proceedings of SC20: International Conference for High Performance Computing, Networking, Storage and Analysis , 2020, pp. 1–12, IEEE

    2020

  24. [32]

    PyTorch: An imperative style, high-performance deep learning library,

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga, et al. , “PyTorch: An imperative style, high-performance deep learning library,” Advances in Neural Information Processing Systems , vol. 32, 2019

    2019

  25. [33]

    Deep graph library: To- wards efficient and scalable deep learning on graphs,

    M. Wang, L. Yu, D. Zheng, Q. Gan, Y . Gai, Z. Ye, M. Li, J. Zhou, Q. Huang, C. Ma, Z. Huang, et al. , “Deep graph library: To- wards efficient and scalable deep learning on graphs,” arXiv preprint arXiv:1909.01315, 2019

    1909 arXiv

  26. [34]

    Exploring Universities with NLP: Topic Modeling and Sentiment Analysis Using Wikipedia Data

    Pei-Hsuan Sung. Exploring Universities with NLP: Topic Modeling and Sentiment Analysis Using Wikipedia Data. TechRxiv. February 07, 2025. DOI: 10.36227/techrxiv.173896964.47108397/v1

    2025

  27. [35]

    Mitigating Challenges in Ethereum’s Proof-of-Stake Consen- sus: Evaluating the Impact of EigenLayer and Lido,

    L. Li, “Mitigating Challenges in Ethereum’s Proof-of-Stake Consen- sus: Evaluating the Impact of EigenLayer and Lido,” *arXiv preprint arXiv:2410.23422*, 2024

    2024 arXiv

  28. [36]

    Spectral graph attention network with fast eigen-approximation,

    H. Chang, Y . Rong, T. Xu, W. Huang, S. Sojoudi, J. Huang, and W. Zhu, “Spectral graph attention network with fast eigen-approximation,” in Proceedings of the 30th ACM International Conference on Information & Knowledge Management , 2021, pp. 2905–2909

    2021

  29. [37]

    S. Li, S. Shi, Y . Xiao, C. Zhang, Y . T. Hou, and W. Lou, ”Bijack: Break- ing bitcoin network with tcp vulnerabilities,” in European Symposium on Research in Computer Security , Springer, 2023, pp. 306–326

    2023

  30. [38]

    UCBlocker: Unwanted call blocking using anonymous authentication,

    C. Du, H. Yu, Y . Xiao, Y . T. Hou, A. D. Keromytis, and W. Lou, “UCBlocker: Unwanted call blocking using anonymous authentication,” in 32nd USENIX Security Symposium (USENIX Security 23) , 2023, pp. 445–462

    2023

  31. [39]

    Aaka: An anti-tracking cellular authentication scheme leveraging anonymous credentials,

    H. Yu, C. Du, Y . Xiao, A. Keromytis, C. Wang, R. Gazda, Y . T. Hou, and W. Lou, “Aaka: An anti-tracking cellular authentication scheme leveraging anonymous credentials,” in Proceedings 2024 Network and Distributed System Security Symposium , 2023

    2024

  32. [40]

    Comprehensive survey on adversarial examples in cyberse- curity: Impacts, challenges, and mitigation strategies,

    L. Li, “Comprehensive survey on adversarial examples in cyberse- curity: Impacts, challenges, and mitigation strategies,” arXiv preprint arXiv:2412.12217, 2024

    2024 arXiv

  33. [41]

    State-of-the-art approaches to enhancing privacy preservation in machine learning datasets: A survey,

    C. Zhang and S. Li, “State-of-the-art approaches to enhancing privacy preservation in machine learning datasets: A survey,” arXiv preprint , arXiv:2404.16847, 2024

    2024 arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.