Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Accelerated Evolving Set Processes for Local PageRank Computation
Authors: Binbin Huang, Luo Luo, Yanghua Xiao, Deqing Yang, Baojian Zhou
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results on real-world graphs validate the efficiency of our methods, demonstrating significant convergence in the early stages. |
| Researcher Affiliation | Academia | Binbin Huang 1 Luo Luo 1,2 Yanghua Xiao 3 Deqing Yang 1,3 Baojian Zhou 1,3 1 School of Data Science, Fudan University, 2 Shanghai Key Laboratory for Contemporary Applied Mathematics, 3 Shanghai Key Laboratory of Data Science, School of Computer Science, Fudan University EMAIL luoluo,shawyh,yangdeqing,EMAIL |
| Pseudocode | Yes | Algorithm 1 AESP(ϵ, α, b, η, G, M) Algorithm 2 AESP-PPR(ϵ, α, s, G, M) Algorithm 3 LOCGD(φt, y(t 1), η, α, b, G) Algorithm 4 LOCAPPR(φt, y(t 1), η, α, b, G) |
| Open Source Code | Yes | Our code is publicly available at https://github.com/Rick7117/aesp-local-pagerank. |
| Open Datasets | Yes | In our main experiments, we evaluate the proposed method on a medium-scale graph com-dblp and four large-scale graphs ogb-mag240m, ogbn-papers100M, com-friendster, and wiki-en21. Table 2: Dataset Statistics G1 as-skitter G2 cit-patent G3 com-dblp G4 com-friendster G5 com-lj G6 com-orkut G7 com-youtube G8 ogb-mag240m G9 ogbl-ppa G10 ogbn-arxiv G11 ogbn-mag G12 ogbn-papers100M G13 ogbn-products G14 ogbn-proteins G15 soc-lj1 G16 soc-pokec G17 sx-stackoverflow G18 wiki-en21 G19 wiki-talk |
| Dataset Splits | No | The paper states, "selecting 50 source nodes s for each α at random," which describes a sampling strategy for problem instances rather than a specific training/test/validation split of the graph datasets themselves. It does not provide explicit dataset splits as typically defined for machine learning tasks. |
| Hardware Specification | Yes | The experiments were conducted using Python 3.10 with Cu Py and Numba libraries on a server with 96 cores, 503Gi B of memory, and four NVIDIA RTX A6000 GPUs with 28GB each. |
| Software Dependencies | No | The paper mentions "Python 3.10 with Cu Py and Numba libraries" but only provides a version number for Python. Specific version numbers for the key libraries Cu Py and Numba are not provided. |
| Experiment Setup | Yes | The damping factor α and convergence threshold ϵ were fixed at α = 0.1 and ϵ = 10^-6 throughout all experiments unless otherwise specified. We fix the precision at ϵ = 10^-7 and vary α from 10^-3 to 10^-1, selecting 50 source nodes s for each α at random. This supplementary investigation further evaluates the performance of AESP-LOCAPPR and AESP-LOCGD under varying initialization approaches (y(t 1) versus momentum-free x(t 1) versus zero-initialization) with fixed parameters α = 0.01 and ϵ = 0.1/n. |