Structural-entropy survey: maps of the field
← benchmark · periodic table · gallery · survey
Three views of the 81 tagged core papers, all generated from the facet tags.
1 · Top 10 periodic-table projections (by coverage)
Each is a different row×col cut; cell colour ∝ #papers. The same corpus, many lenses.
task × output (81 papers)
| partition | tree | pooled_graph | embedding | rewired | policy | score |
|---|
| theory | 8 | 3 | 1 | 1 | 3 | 4 | 9 |
|---|
| community_detection | 19 | 15 | 2 | 4 | 8 | 9 | 7 |
|---|
| hierarchical_clustering | 29 | 28 | 5 | 8 | 18 | 16 | 11 |
|---|
| graph_pooling | 3 | 3 | 3 | 1 | 3 | · | 1 |
|---|
| structure_learning | 4 | 9 | · | 6 | 11 | 6 | 7 |
|---|
| contrastive_ssl | 1 | 2 | · | 2 | 2 | 2 | 2 |
|---|
| rl_abstraction | 6 | 6 | · | 4 | 4 | 8 | 4 |
|---|
| application | 22 | 22 | 2 | 6 | 9 | 14 | 14 |
|---|
graph_type × output (81 papers)
| partition | tree | pooled_graph | embedding | rewired | policy | score |
|---|
| simple | 16 | 17 | · | 9 | 11 | 13 | 15 |
|---|
| weighted | 10 | 9 | 2 | · | 4 | 4 | 7 |
|---|
| directed | 4 | 3 | · | 3 | 2 | 5 | 5 |
|---|
| bipartite | 1 | 1 | 1 | 1 | 1 | 1 | · |
|---|
| attributed | 16 | 15 | 4 | 8 | 12 | 7 | 4 |
|---|
| heterogeneous | 10 | 11 | 2 | 3 | 10 | 8 | 6 |
|---|
| hypergraph | 1 | 1 | 1 | 1 | 1 | 1 | · |
|---|
| dynamic | 24 | 22 | 2 | 6 | 13 | 18 | 17 |
|---|
task × se_role (81 papers)
| objective | regularizer | prior | metric | feature | robustness |
|---|
| theory | 9 | 1 | 3 | 8 | 2 | 1 |
|---|
| community_detection | 19 | 1 | 8 | 14 | 7 | 7 |
|---|
| hierarchical_clustering | 25 | 3 | 21 | 22 | 17 | 10 |
|---|
| graph_pooling | 1 | · | 3 | 3 | 2 | 1 |
|---|
| structure_learning | 4 | 2 | 11 | 9 | 7 | 5 |
|---|
| contrastive_ssl | 1 | · | 2 | 2 | 2 | 2 |
|---|
| rl_abstraction | 3 | 2 | 8 | 6 | 5 | 3 |
|---|
| application | 10 | 1 | 15 | 17 | 16 | 14 |
|---|
se_role × output (81 papers)
| partition | tree | pooled_graph | embedding | rewired | policy | score |
|---|
| objective | 38 | 34 | 4 | 12 | 21 | 24 | 26 |
|---|
| regularizer | 3 | 6 | · | 4 | 4 | 5 | 5 |
|---|
| prior | 35 | 40 | 4 | 20 | 29 | 28 | 26 |
|---|
| metric | 47 | 45 | 6 | 18 | 31 | 33 | 40 |
|---|
| feature | 30 | 34 | 6 | 24 | 26 | 23 | 19 |
|---|
| robustness | 25 | 26 | 3 | 12 | 16 | 19 | 25 |
|---|
graph_type × se_role (81 papers)
| objective | regularizer | prior | metric | feature | robustness |
|---|
| simple | 18 | 2 | 17 | 19 | 11 | 9 |
|---|
| weighted | 7 | · | 4 | 8 | 5 | 5 |
|---|
| directed | 4 | 1 | 4 | 4 | 4 | 1 |
|---|
| bipartite | · | · | · | · | 1 | · |
|---|
| attributed | 7 | 2 | 13 | 14 | 14 | 9 |
|---|
| heterogeneous | 5 | 2 | 9 | 10 | 8 | 5 |
|---|
| hypergraph | · | · | · | · | 1 | · |
|---|
| dynamic | 15 | 3 | 14 | 19 | 14 | 13 |
|---|
graph_type × task (81 papers)
| theory | community_detection | hierarchical_clustering | graph_pooling | structure_learning | contrastive_ssl | rl_abstraction | application |
|---|
| simple | 3 | 7 | 9 | · | 5 | 1 | 2 | 9 |
|---|
| weighted | 1 | 5 | 7 | 1 | · | · | · | 3 |
|---|
| directed | 2 | 1 | 3 | · | 2 | · | 1 | · |
|---|
| bipartite | · | · | · | · | · | · | · | 1 |
|---|
| attributed | · | 3 | 7 | 2 | 3 | 1 | · | 8 |
|---|
| heterogeneous | 1 | · | 2 | · | 2 | · | 1 | 10 |
|---|
| hypergraph | · | · | · | · | · | · | · | 1 |
|---|
| dynamic | 3 | 6 | 12 | · | 1 | · | 5 | 12 |
|---|
supervision × output (81 papers)
| partition | tree | pooled_graph | embedding | rewired | policy | score |
|---|
| unsupervised | 54 | 45 | 6 | 14 | 23 | 32 | 33 |
|---|
| self_supervised | 17 | 18 | 4 | 8 | 13 | 12 | 13 |
|---|
| supervised | 17 | 22 | 3 | 14 | 20 | 14 | 14 |
|---|
| semi_supervised | 2 | 2 | · | · | · | 1 | 1 |
|---|
graph_type × supervision (81 papers)
| unsupervised | self_supervised | supervised | semi_supervised |
|---|
| simple | 21 | 6 | 11 | 2 |
|---|
| weighted | 10 | 1 | 2 | · |
|---|
| directed | 3 | 2 | 2 | · |
|---|
| bipartite | 1 | 1 | · | · |
|---|
| attributed | 11 | 8 | 8 | · |
|---|
| heterogeneous | 11 | 7 | 6 | · |
|---|
| hypergraph | 1 | 1 | · | · |
|---|
| dynamic | 24 | 8 | 6 | · |
|---|
task × supervision (81 papers)
| unsupervised | self_supervised | supervised | semi_supervised |
|---|
| theory | 9 | 1 | 1 | · |
|---|
| community_detection | 19 | 2 | 1 | 2 |
|---|
| hierarchical_clustering | 25 | 6 | 9 | 2 |
|---|
| graph_pooling | 1 | · | 3 | · |
|---|
| structure_learning | · | 4 | 11 | · |
|---|
| contrastive_ssl | · | 2 | · | · |
|---|
| rl_abstraction | 5 | 1 | 2 | · |
|---|
| application | 29 | 13 | 11 | · |
|---|
se_role × supervision (81 papers)
| unsupervised | self_supervised | supervised | semi_supervised |
|---|
| objective | 40 | 11 | 13 | 2 |
|---|
| regularizer | 4 | 1 | 5 | 1 |
|---|
| prior | 32 | 14 | 26 | 2 |
|---|
| metric | 44 | 15 | 21 | 1 |
|---|
| feature | 28 | 18 | 17 | · |
|---|
| robustness | 25 | 13 | 13 | · |
|---|
2 · Method genealogy
Who builds on whom (curated lineage; edge = extends / is the deep version of).
Li & Pan 2016 — k-D structural entropy (the source)- deDoc 2018 — greedy 2D-SE minimization (Hi-C)
- SuperTAD 2021 — DP encoding tree (Hi-C TADs)
- SuperTAD2 2024
- deTOKI 2021 — TAD-insulation
- deDoc2 2023 — DP whole-map 2D-SE (Hi-C)
- pySEAT 2022 — single-cell hierarchy
- HCSE 2021 — hierarchical clustering via SE
- BBM — balanced binary merge (encoding-tree baseline)
- se_hier (this lib) — exact-guarded tree refinement
- HCSE-graphs 2025 (UAI) — graph version
- CoDeSEG 2025 — SE-game community detection
- se_louvain (this lib) — multilevel 2D-SE
- DeSE — attributed deep SE clustering
- LSENet — Lorentz/hyperbolic SE clustering
- se_gnn (this lib) — soft 2D-SE + GCN + Sinkhorn
- SEP 2022 (ICML) — SE-guided graph pooling
- SE-GSL — SE graph structure learning
- SEGA — SE-anchored graph contrastive learning
- HiTIN — hierarchical text classification
- SI2E — SE intrinsic exploration (RL)
- SIDM — SE hierarchical decision making
- SISA — SE state abstraction
- Extensions of the definition
- Flow-based directed SE 2019
- Multi-Relational SE 2024 (UAI)
- Incremental SE 2024 (dynamic graphs)
- Hypergraph SE 2025
- EDEN 2025 — directed-graph deep SE
- Social / event / bot apps
- Hierarchical+Incremental SE events 2024
- UnDBot — social bot detection
- SEIN 2026 — open-world events
- LLM / uncertainty (emerging)
- SeSE — structural uncertainty / hallucination
- RAG-SEDE — retrieval structuring
3 · Phase diagram — when each direction emerged
x = year, lane = research direction, dot = paper, colour = graph type.
●simple ●weighted ●directed ●attributed ●heterogeneous ●dynamic ●bipartite