The six matrices in this directory are unchanged Matrix Market text files, compressed with gzip, downloaded on 2026-09-23 from the University of Florida mirror of the SuiteSparse Matrix Collection. The Les Miserables vertex-name file is also unchanged. Each matrix retains its original metadata header.
| Source | Original authors | Editors |
|---|---|---|
| HB/dwt_66 | G. Everstine, D. Taylor | I. Duff, R. Grimes, J. Lewis |
| Newman/lesmis | D. Knuth | M. Newman |
| HB/dwt_307 | G. Everstine, D. Taylor | I. Duff, R. Grimes, J. Lewis |
| HB/494_bus | D. Tylavsky | I. Duff, R. Grimes, J. Lewis |
| HB/dwt_1005 | G. Everstine, D. Taylor | I. Duff, R. Grimes, J. Lewis |
| HB/1138_bus | D. Tylavsky | I. Duff, R. Grimes, J. Lewis |
Source archives:
https://www.cise.ufl.edu/research/sparse/MM/GROUP/NAME.tar.gz,
using the group/name in the table. The current Texas A&M HTTPS file
server was unreachable during this import; the collection’s original
Florida mirror was used.
SuiteSparse distributes the matrices under Creative Commons
Attribution 4.0 International, as stated in the collection’s license policy.
This license applies to these matrices, their metadata and the derived
graphs returned by suitesparse.graph(), separately from
grip’s software license.
Please credit the original authors above and cite Davis, T. A. and Hu, Y. (2011), The University of Florida Sparse Matrix Collection, ACM Transactions on Mathematical Software 38(1), Article 1, https://doi.org/10.1145/2049662.2049663. For Les Miserables, also cite Knuth, D. E. (1993), The Stanford GraphBase: A Platform for Combinatorial Computing, Addison-Wesley. The complete source-specific notes remain in each matrix header.
The examples were selected from Zheng, J. X., Pawar, S. and Goodman, D. F. M. (2018), Graph Drawing by Stochastic Gradient Descent, https://doi.org/10.1109/TVCG.2018.2859997.
data-raw/original_six.R reconstructs the six examples
offline from these files as part of
data-raw/build-suitesparse.py. Derived names end in
_unweighted_graph. The conversion removes diagonal entries,
discards explicit zeros, combines symmetric/duplicate edges, and assigns
unit traversal lengths. All vertices and original row indices are
retained. All six graphs are connected; no component was dropped or
repaired. Original matrix values are preserved in the archived files,
not treated as distances. The source names and per-file checksums are
retained in each graph’s metadata.