Graph Dataset for "Jet: Multilevel Graph Partitioning on Graphics Processing Units"
Bibliographic record
Abstract
*This README file was made on 2024-05-08 by Michael S. Gilbert* This dataset is a compressed archive of 65 graphs in the METIS graph file format. These graphs have been used to test and measure the capabilities of Jet, a novel hiqh-quality, GPU-parallel, k-way refinement algorithm (https://github.com/sandialabs/Jet-Partitioner). Methodological information can be found in our related publication below. ## Author Information: * Michael S. Gilbert, msg5334@psu.edu. Pennsylvania State University, University Park, USA. * Kamesh Madduri, madduri@psu.edu. Pennsylvania State University, University Park, USA. * Erik G. Boman, egboman@sandia.gov. Sandia National Laboratories, Albuquerque, USA. * Sivasankaran Rajamanickam, srajama@sandia.gov. Sandia National Laboratories, Albuquerque, USA. ## Funders and Sponsors Sandia National Laboratories is a multimission laboratory managed and operated by National Technology and Engineering Solutions of Sandia, LLC., a wholly owned subsidiary of Honeywell International, Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA-0003525. This research was supported by the Exascale Computing Project (17-SC-20-SC), a collaborative effort of the U.S. Department of Energy Office of Science and the National Nuclear Security Administration. ## Sharing/Access Information ### License No copyright retained - U.S. Public Domain. ### Recommended citation for this data Gilbert, Michael; Madduri, Kamesh; Boman, Erik; Rajamanickam, Sivasankaran (2024). Graph Dataset for "Jet: Multilevel Graph Partitioning on Graphics Processing Units" [Data set]. Scholarsphere. https://doi.org/10.26207/pffm-mc36. ### Related publications Gilbert, Michael S., et al. Jet: Multilevel Graph Partitioning on Graphics Processing Units. 2023. DOI.org (Datacite), https://doi.org/10.48550/ARXIV.2304.13194. ## Data & File Overview ### File list The compression utility used to compress the graphs is "xz". The cksum output for the compressed archive is "731682270 7880707444 graphs.tar.xz". We preprocessed all graphs by performing the following steps: remove self-loops, convert all directed edges to undirected edges, remove duplicate edges, and extract the largest connected component. The following are the sources for each graph: 1. Originally created graphs: * grid_3.graph, 2000x4000 rectangular mesh (grid) * cube_2.graph, 200x200x200 cubic mesh (cube) 2. Suitesparse graph repository - T. A. Davis and Y. Hu, The University of Florida sparse matrix collection, ACM Trans. on Mathematical Software, 38 (2011). https://dl.acm.org/doi/10.1145/2049662.2049663. We included all graphs (except the mawi graphs) with at least 50 million nonzeroes but less than 750 million nonzeroes: * Bump_2911.graph * Cube\_Coup\_dt0.graph * Cube\_Coup\_dt6.graph * Flan_1565.graph * Geo_1438.graph * HV15R.graph * Hook_1498.graph * Long\_Coup\_dt0.graph * Long\_Coup\_dt6.graph * ML_Geer.graph * Queen_4147.graph * Serena.graph * af_shell10.graph * arabic-2005.graph * audikw_1.graph * cage15.graph * channel-500x100x100-b050.graph * circuit5M.graph * com-LiveJournal.graph * com-Orkut.graph * delaunay_n23.graph * delaunay_n24.graph * dielFilterV3real.graph * europe_osm.graph * hollywood-2009.graph * hugebubbles-00000.graph * hugebubbles-00010.graph * hugebubbles-00020.graph * indochina-2004.graph * kmer_A2a.graph * kmer_P1a.graph * kmer_U1a.graph * kmer_V1r.graph * kmer_V2a.graph * kron_g500-logn20.graph * kron_g500-logn21.graph * ljournal-2008.graph * mycielskian17.graph * mycielskian18.graph * nlpkkt120.graph * nlpkkt160.graph * nlpkkt200.graph * rgg\_n\_2\_22\_s0.graph * rgg\_n\_2\_23\_s0.graph * rgg\_n\_2\_24\_s0.graph * road_usa.graph * soc-LiveJournal1.graph * soc-Pokec.graph * stokes.graph * uk-2002.graph * vas\_stokes\_2M.graph * vas\_stokes\_4M.graph * vsp\_bcsstk30\_500sep\_10in\_1Kout.graph * vsp\_vibrobox\_scagr7-2c\_rlfddd.graph * wb-edu.graph 3. Miscellaneous graphs from the Open Graph Benchmark - W. Hu, M. Fey, M. Zitnik, Y. Dong, H. Ren, B. Liu, M. Catasta, and J. Leskovec, Open Graph Benchmark: Datasets for machine learning on graphs, in Proc. Annual Conf. on Neural Inf. Proc. Systems, 2020. https://proceedings.neurips.cc/paper/2020/hash/fb60d411a5c5b72b2e7d3527cfc84fd0-Abstract.html. * citation.graph * products.graph * ppa.graph 4. Social network graphs from the Laboratory for Web Algorithmics - P. Boldi, M. Rosa, M. Santini, and S. Vigna, Layered label propagation: A multiresolution coordinate-free ordering for compressing social networks, in Proc. 20th Int’l. Conf. on World Wide Web (WWW), 2011. https://dl.acm.org/doi/10.1145/1963405.1963488. * dblp-2010.graph * amazon-2008.graph * hollywood-2011.graph * enwiki-2021.graph 5. Walshaw Graph Benchmark - A. J. Soper, C. Walshaw, and M. Cross, A combined evolutionary search and multilevel optimisation approach to graph-partitioning, Journal of Global Optimization, 29 (2004), pp. 225–241, https://api.semanticscholar.org/CorpusID:6904637. * fe_rotor.graph ## Computational Dependencies These graphs may be used with any software which supports the METIS graph file format. The graphs were later checked and verified using the `graphck` utility from METIS on a Kubuntu 22.04 x64 (5.15.0-101-generic) desktop, using cmake version 3.22.1, g++ (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0, and METIS version 5.1.0.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.004 |
| Meta-epidemiology (narrow) | 0.003 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.002 | 0.004 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.002 | 0.001 |
| Open science | 0.003 | 0.002 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.087 | 0.079 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".