Published April 2, 2026 | Version v1
Dataset Open

Benchmark Sets, Training Data and Results for "Engineering Learned Heuristics to Improve Clustering for Multilevel Graph Partitioning"

Authors/Creators

  • 1. ROR icon Karlsruhe Institute of Technology
  • 2. ROR icon Heidelberg University

Description

  • set_a.zip: benchmark set of 118 small to medium graphs. All graphs are unweighted and in Metis format
  • set_b.zip: subset of Set A, consisting of 69 instances, used as a training set for our models
  • set_b1-20.zip: subset of Set B, consisting of 20 instances with high mean edge label values
  • set_b21-45.zip: subset of Set B, consisting of 25 instances with medium mean edge label values
  • set_b46-69.zip: subset of Set B, consisting of 24 instances with low mean edge label values
  • features_set_b.zip: input features that we used for our models for all graphs of Set B
  • labels_set_b.zip: training edge labels for all graphs of Set B
  • benchmark_results.zip: experimental results for all evaluated graph partitioning algorithms

Files

benchmark_results.zip

Files (37.5 GB)

Name Size
md5:8812a8addd84b151112afb15c6825614
2.5 MB Preview Download
md5:c8a9816ec896c251f31b29c93cd5d197
27.2 GB Preview Download
md5:5ea3e14c42dfe5d60073df5f66b58184
3.3 GB Preview Download
md5:94606409c10bced915890cf6e6fda213
3.4 GB Preview Download
md5:32e4408cc80a22a609304718c63875ea
1.8 GB Preview Download
md5:3ef66693f29b3b544a3582423bd8ccab
857.7 MB Preview Download
md5:33ba867dcf90d048a784a6339fc54316
337.4 MB Preview Download
md5:7de231e52cf2e11fa0ec668d129aae8b
559.4 MB Preview Download