Add model information to mwis2024 page

vislearn · Aug 6, 2024 · e1f6a86 · e1f6a86
1 parent dab9af2
commit e1f6a86
Showing 1 changed file with 194 additions and 1 deletion.
diff --git a/publications/mwis2024.md b/publications/mwis2024.md
@@ -2,4 +2,197 @@
 permalink: /mwis2024/
 ---
 
-Content will soon get published.
+## A Bregman-Sinkhorn Algorithm for the Maximum Weight Independent Set Problem (arXiv 2024 Preprint)
+
+This pages is dedicated to our [arXiv 2024 preprint “A Bregman-Sinkhorn Algorithm for the Maximum Weight Independent Set Problem”][arxiv2024].
+
+### Citation
+
+S. Haller, B. Savchynskyy<br>
+**“A Bregman-Sinkhorn Algorithm for the Maximum Weight Independent Set Problem”**<br>
+*arXiv 2408.02086 Preprint 2024* [[PDF][arxiv2024]]
+
+BibTeX entry:
+
+```
+@misc{haller2024mwis,
+  title={A Bregman-Sinkhorn Algorithm for the Maximum Weight Independent Set Problem},
+  author={Stefan Haller and Bogdan Savchynskyy},
+  year={2024},
+  eprint={2408.02086},
+  archivePrefix={arXiv},
+  primaryClass={math.OC},
+  url={https://arxiv.org/abs/2408.02086},
+}
+```
+
+[arxiv2024]: https://arxiv.org/pdf/2408.02086
+
+## Problem Instances / Model Files
+
+### JSON Format
+
+Our solver uses a simple JSON graph format:
+
+```jsonc
+{
+  "nodes": [10.0, 20.0, 30.0], /* Costs for each node */
+  "cliques": [[0, 1, 3]], /* Clique formation (indices into nodes array) */
+  "constant": 0.0, /* Constant that is added to the objective unconditionally */
+  "edges": [[0, 1], [0, 2], [1, 2]], /* Pairwise edge representation (indices into nodes array) */
+}
+```
+
+Donwload JSON model files: [mwis2024_json.tar.zst]
+
+### Metis Format
+
+For the specification of the metis format refer to the [Metis Software Package Manual][metis_format].
+However, note that only edge graph representation is supported, i.e., the edges for all cliques are enumerated in its entirety.
+
+Download Metis model files: [mwis2024_metis.tar.zst]
+
+### Dataset Groups
+
+In the paper we subdivide the large and relatively diverse collection of AVR datasets into three groups (LARGE, MEDIUM and SMALL).
+The groups are selected based on the number of nodes in the problem instances.
+More information are presented in the paper.
+
+<details>
+<summary>Group mapping in JSON format</summary>
+
+```json
+{
+  "AVR-LARGE": [
+    "AVR_000",
+    "AVR_001",
+    "AVR_002",
+    "AVR_003",
+    "AVR_004",
+    "AVR_005",
+    "AVR_006",
+    "AVR_007",
+    "AVR_008",
+    "AVR_009",
+    "AVR_010",
+    "AVR_011",
+    "AVR_012",
+    "AVR_013",
+    "AVR_014",
+    "AVR_015",
+    "AVR_016"
+  ],
+  "AVR-MEDIUM": [
+    "AVR_017",
+    "AVR_018",
+    "AVR_020",
+    "AVR_021",
+    "AVR_022",
+    "AVR_023",
+    "AVR_025",
+    "AVR_026",
+    "AVR_028",
+    "AVR_029",
+    "AVR_031",
+    "AVR_032",
+    "AVR_033",
+    "AVR_035",
+    "AVR_036",
+    "AVR_037"
+  ],
+  "AVR-SMALL": [
+    "AVR_019",
+    "AVR_024",
+    "AVR_027",
+    "AVR_030",
+    "AVR_034"
+  ],
+  "MSCD": [
+    "MSCD_000",
+    "MSCD_001",
+    "MSCD_002",
+    "MSCD_003",
+    "MSCD_004",
+    "MSCD_005",
+    "MSCD_006",
+    "MSCD_007",
+    "MSCD_008",
+    "MSCD_009",
+    "MSCD_010",
+    "MSCD_011",
+    "MSCD_012",
+    "MSCD_013",
+    "MSCD_014",
+    "MSCD_015",
+    "MSCD_016",
+    "MSCD_017",
+    "MSCD_018",
+    "MSCD_019",
+    "MSCD_020"
+  ]
+}
+```
+</details>
+
+## Origin of AVR datasets
+
+The AVR problem instances have been original obtained from the [Amazon MWIS VR Dataset] page.
+The problem instances have been used as is.
+We converted the text file representation into our JSON clique graph format as well as into the Metis edge graph format.
+
+<details>
+<summary>Mapping of instance number to original instance name as JSON</summary>
+
+```json
+{
+  "AVR_000": "CR-S-L-1",
+  "AVR_001": "CR-S-L-2",
+  "AVR_002": "CR-S-L-4",
+  "AVR_003": "CR-S-L-6",
+  "AVR_004": "CR-S-L-7",
+  "AVR_005": "CR-T-C-1",
+  "AVR_006": "CR-T-C-2",
+  "AVR_007": "CR-T-D-4",
+  "AVR_008": "CR-T-D-6",
+  "AVR_009": "CR-T-D-7",
+  "AVR_010": "CW-S-L-1",
+  "AVR_011": "CW-S-L-2",
+  "AVR_012": "CW-S-L-4",
+  "AVR_013": "CW-S-L-6",
+  "AVR_014": "CW-S-L-7",
+  "AVR_015": "CW-T-C-1",
+  "AVR_016": "CW-T-C-2",
+  "AVR_017": "CW-T-D-4",
+  "AVR_018": "CW-T-D-6",
+  "AVR_019": "MR-D-01",
+  "AVR_020": "MR-D-03",
+  "AVR_021": "MR-D-05",
+  "AVR_022": "MR-D-FN",
+  "AVR_023": "MR-W-FN",
+  "AVR_024": "MT-D-01",
+  "AVR_025": "MT-D-200",
+  "AVR_026": "MT-D-FN",
+  "AVR_027": "MT-W-01",
+  "AVR_028": "MT-W-200",
+  "AVR_029": "MT-W-FN",
+  "AVR_030": "MW-D-01",
+  "AVR_031": "MW-D-20",
+  "AVR_032": "MW-D-40",
+  "AVR_033": "MW-D-FN",
+  "AVR_034": "MW-W-01",
+  "AVR_035": "MW-W-05",
+  "AVR_036": "MW-W-10",
+  "AVR_037": "MW-W-FN"
+}
+```
+</details>
+
+## Source Code for Publication
+
+Coming soon.
+
+
+[Amazon MWIS VR Dataset]: https://registry.opendata.aws/mwis-vr-instances/
+
+[mwis2024_json.tar.zst]: https://www.stha.de/shares/mwis2024/mwis2024_json.tar.zst
+[mwis2024_metis.tar.zst]: https://www.stha.de/shares/mwis2024/mwis2024_metis.tar.zst