---
title: 2D Structural Entropy Minimization
url: https://www.emergentmind.com/topics/two-dimensional-structural-entropy-minimization
type: topic
---

# 2D Structural Entropy Minimization

Searching arXiv for recent and foundational papers on two-dimensional structural entropy minimization and related usage.
arXiv search query: "structural entropy minimization encoding tree two-dimensional graph partitioning"
Two-dimensional structural entropy minimization is, in its most common contemporary usage, a graph-partitioning principle in which one seeks a height-2 encoding tree that minimizes the information needed to describe graph structure under a two-level abstraction: root, modules, and individual vertices. In this formulation, the intermediate nodes of the encoding tree are the sought communities, regions, events, sentence groups, or sub-dialogues, and minimization acts as an unsupervised criterion for discovering them without pre-specifying their number [2312.11891][2309.01899][2507.23400][2508.05023]. The same phrase also appears in mathematically distinct literatures on entropy for embedded surfaces and on entropy from recurring structural patterns in two-dimensional liquids; those usages concern different entropy functionals and should not be conflated with the encoding-tree formalism [1509.06238][2605.29945].

## 1. Formal framework

In the encoding-tree formalism, an encoding tree \(\mathcal{T}\) of a graph \(G=(\mathcal{V},\mathcal{E})\) is a hierarchical partition: the root \(\lambda\) corresponds to all vertices, each internal node \(\alpha\) corresponds to a subset \(T_\alpha\subseteq \mathcal{V}\), and leaves correspond to singleton graph nodes. A standard definition used across recent papers is
\[
\mathcal{H}^{\mathcal{T}}(G)
=
-\sum_{\alpha\in\mathcal{T},\,\alpha\neq\lambda}
\frac{g_\alpha}{vol(\lambda)}
\log
\frac{vol(\alpha)}{vol(\alpha^-)},
\]
where \(g_\alpha\) is the cut weight of \(T_\alpha\), \(vol(\alpha)\) is the weighted volume of \(T_\alpha\), and \(\alpha^-\) is the parent of \(\alpha\). The \(d\)-dimensional structural entropy is then
\[
\mathcal{H}^{(d)}(G)=\min_{\forall\mathcal{T}:h(\mathcal{T})=d}\{\mathcal{H}^{\mathcal{T}}(G)\}.
\]
For two-dimensional structural entropy, \(h(\mathcal{T})=2\): the root is the full graph, the height-1 nodes are modules, and the leaves are individual vertices [2312.11891][2508.05023].

A closely related partition-level presentation is used in image segmentation. For a partition \(\mathcal{P}=\{X_1,\dots,X_L\}\) of an undirected weighted graph \(G=(V,E,\mathbf W)\), the structural entropy is written as
\[
\mathcal{H}^{\mathcal{P}}(G)
=
-\sum_{X\in\mathcal P}\sum_{v_i\in X}\frac{g_i}{vol(G)}\log_2\frac{d_i}{vol(X)}
-\sum_{X\in\mathcal P}\frac{cut(X)}{vol(G)}\log_2\frac{vol(X)}{vol(G)}.
\]
This is explicitly identified as “the two-dimensional SE, i.e., SE of a graph by a partition,” so the “two-dimensional” qualifier there refers to the two-level coding structure rather than to Euclidean image coordinates [2309.01899].

Across these formulations, the operative meaning is stable. Two-dimensional structural entropy measures the coding cost of first locating a module and then a node within that module. Minimization therefore favors partitions with strong internal coherence and weak external connectivity. This suggests why the method repeatedly appears in unsupervised clustering settings: it aligns graph compression, community structure, and adaptive partition discovery [2312.11891][2207.12653].

## 2. Optimization principles and algorithmic forms

The basic optimization pattern is greedy agglomeration on the height-1 nodes of a 2D encoding tree. In the event-detection formulation, one initializes each message as a singleton cluster, computes \(\Delta \text{SE}\) for every pair of current height-1 nodes, chooses the pair with the most negative \(\Delta \text{SE}\), merges it, and stops when no pair gives \(\Delta \text{SE}<0\). The resulting partition is the final event decomposition, and the number of clusters is determined implicitly by the stopping condition rather than supplied externally [2312.11891].

Several papers then customize this baseline. HISEvent replaces global cubic-time greedy merging with a hierarchical procedure: it chunks the current cluster list into subsets of size \(n\), induces subgraphs, initializes a local height-2 tree using the already detected clusters as height-1 nodes, runs vanilla 2D SE minimization locally, and doubles \(n\) if a round stagnates. The stated purpose is to retain the 2D SE objective while making it practical on large social message graphs; the paper reports \(O(|\mathcal V|^3)\) for vanilla 2D SE minimization and \(O(n^3)\) for the hierarchical version [2312.11891].

SLED uses a different refinement strategy. It first performs a merge stage, greedily selecting the pair \((X,Y)\) with the largest positive entropy decrease \(\Delta \mathcal H^{\mathcal M}_{X,Y}\), and then performs a refinement stage in which individual nodes are removed from their current module and inserted into another module according to the objective
\[
\arg\max_Y\left\{\Delta \mathcal H^{\mathcal R}_{X,v_i}-\Delta \mathcal H^{\mathcal I}_{Y,v_i}\right\}.
\]
The paper explicitly argues monotonicity: every accepted merge decreases \(\mathcal H^{\mathcal P}(G)\), and each node reassignment decreases or preserves it, so the procedure converges when no further improvement exists [2309.01899].

Dynamic-graph work turns the static optimizer into an incremental one. Incre-2dSE maintains Structural Data, Structural Expressions, and Adjustment objects, then updates the entropy through an incremental formula rather than recomputing from scratch:
\[
H^{\mathcal T'}(G')
=
-\frac{1}{2m+2n}
\bigl[
\hat S'_N+\hat S'_C+\hat S'_G\log(2m+2n)
\bigr].
\]
Its stronger node-shifting strategy defines an optimal preference community (OPC) for a node and greedily moves involved nodes toward entropy-reducing communities, thereby incrementally optimizing the partition toward lower 2D structural entropy over time [2207.12653].

A further variant appears in privacy-preserving event detection. ADP-SEMEvent constructs “optimal subgraphs” by starting from the largest edge in a coarsened super-graph, expanding the subgraph with the node of highest cut weight until size \(q\), then applying vanilla 2D SE minimization inside each subgraph. The novelty there is not a new entropy formula but a decomposition heuristic meant to preserve strongly connected partitions in noisy privatized graphs [2407.18274].

## 3. Canonical graph constructions and application targets

The objects being partitioned vary widely, but the graph-theoretic role of 2D SE is stable: vertices encode atomic units, weighted edges encode task-specific affinity, and the height-1 nodes of the optimal 2D tree are the task outputs.

| Domain | Graph nodes and edges | Height-1 clusters represent |
|---|---|---|
| Social event detection | Messages; edges from shared social attributes and semantic similarity | Detected events |
| Skin lesion segmentation | Superpixels; edges from spatial proximity and color similarity | Homogeneous image regions |
| Multi-document summarization | Sentences; fused semantic and discourse relations | Coherent sentence groups |
| Dialogue ASQ extraction | Utterances; reply-thread-aware semantic links | Semantically independent sub-dialogues |

In HISEvent, the graph is built from messages, common attributes such as sender, mentions, hashtags, and named entities, plus semantic edges chosen through incremental 1D SE minimization; 2D SE then partitions the augmented graph into events. The paper states that this stage “eliminates the noise and reveals the essential 2nd-order (cluster-wise) structure underneath the raw graph with no prior knowledge of the number of event clusters,” and reports closed-set examples of Event2012 ARI \(0.50\), AMI \(0.81\), and Event2018 ARI \(0.44\), AMI \(0.66\) [2312.11891].

In SLED, a 2D dermoscopic image is transformed into a superpixel graph. Nodes are SLIC superpixels, edges connect superpixels whose spatial distance is smaller than a threshold \(r\), weights are
\[
\mathbf W_{i,j}=\exp\left(\frac{-d(v_i,v_j)^2}{\sigma_i\sigma_j}\right),
\]
and the graph is sparsified into a \(K\)-nearest-neighbor graph. Minimizing 2D SE yields an adaptive partition into homogeneous regions, after which between-class variance maximization and a multi-scale isolation-forest mechanism separate lesion from healthy skin. The paper reports, for example, on ISIC 2016: SLED\(_{SS}\) DI \(81.61\), JA \(72.21\), and SLED\(_{MS}\) DI \(84.33\), JA \(76.02\) [2309.01899].

In MRGSEM-Sum, sentence nodes are connected by semantic and discourse relations. Because 2D SE “cannot be directly applied to multi-relational graphs,” the relation layers are fused into an integrated weighted graph by elementwise max, and 2D SE minimization is then used to obtain root-level sentence clusters for downstream compression. The paper’s ablation against spectral clustering reports substantial degradation when 2D SE clustering is removed; on Multi-News, MRGSEM-Sum obtains \(43.21/14.66/16.99\) versus \(37.86/11.65/13.05\) for the variant without 2D SE clustering [2507.23400].

In dialogue aspect-based sentiment quadruple extraction, DSEM builds an undirected weighted utterance graph in which edges follow reply relations and thread-reachable utterance pairs, with weights
\[
w_{ij}=\max(\cosine(v_i,v_j),0).
\]
The optimal 2D encoding tree defines sub-dialogues; quadruple matching is then restricted to entities inside the same sub-dialogue. The ablations report that replacing DSEM with K-means degrades EN Micro-F1 from \(42.40\) to \(40.58\) and ZH Micro-F1 from \(44.61\) to \(43.05\) [2508.05023].

## 4. Scalability, privacy, and dynamic maintenance

A recurring theme in the literature is that the 2D SE objective is attractive precisely because it does not require a predefined cluster count, but naïve optimization is expensive. The main engineering innovations therefore concern search-space restriction, incremental maintenance, and robustness to graph noise.

HISEvent’s hierarchical 2D SE minimization is one such response. The paper emphasizes that vanilla 2D SE minimization is cubic in \(|\mathcal V|\) and reports that it can take more than 5 days on a dense Event2012 block and more than 10 days in one closed-set ablation, whereas the hierarchical version reduces runtime by \(>97\%\) on \(M_1\) while preserving performance. It also notes that the split subgraphs can be processed concurrently, making the method parallelizable [2312.11891].

ADP-SEMEvent addresses a different failure mode: the message graph is already noisy because it has been privatized. Its first stage uses an adaptive differential privacy mechanism to construct a private message graph, with 1D structural entropy used to choose \(k\)-NN similarity edges; its second stage uses “a novel 2-dimensional structural entropy minimization algorithm based on optimal subgraphs” to recover events from the noisy graph. The clustering stage is explicitly framed as post-processing of the released private graph, so it “does not compromise differential privacy” [2407.18274].

Incre-2dSE addresses evolving graphs rather than privatized ones. It proves that, under the naïve adjustment strategy, the Local Difference between the Global Invariant approximation and the exact updated entropy satisfies
\[
\Delta L = O\left(\frac{\log m}{m}\right),
\]
and it gives an \(O(n)\) AIUA update procedure for incremental size \(n\). The node-shifting strategy then augments this measurement framework with entropy-guided partition repair, and the experiments report large speedups over offline recomputation, such as \(140.93\times\) on SBM-Hawkes versus Infomap and \(77.81\times\) on DBLP versus Infomap [2207.12653].

MRGSEM-Sum provides a summarization-specific batching analogue. Its 2D SE clustering algorithm groups current clusters into subsets of size \(n\), optimizes locally, and doubles \(n\) when no updates occur. The paper does not give a formal asymptotic complexity theorem, but the batching logic is explicitly presented as the mechanism that makes sentence-level SE clustering tractable beyond a monolithic full-graph search [2507.23400].

## 5. Distinct non-graph usages of entropy minimization on two-dimensional structures

The phrase is not uniform across the broader arXiv literature. In geometric analysis, “entropy” refers to the Colding–Minicozzi entropy for hypersurfaces or self-shrinkers, defined from the Gaussian area or \(F\)-functional rather than from an encoding tree. For a smooth embedded surface \(\Sigma\subset\mathbb R^3\),
\[
\lambda(\Sigma)=\sup_{x_0\in\mathbb R^3,\;t_0>0}F_{x_0,t_0}(\Sigma),
\]
and Ketover and Zhou prove the sharp genus-zero statement
\[
\lambda(\Sigma)\ge \lambda(\mathbb S^2)
\]
for every smooth closed embedded \(2\)-sphere in \(\mathbb R^3\). Their method uses a canonical four-parameter translate-and-dilate family, Gaussian min-max theory, and a degree argument ruling out degeneration to a plane [1509.06238].

A related higher-codimension paper studies smooth, complete, embedded, boundaryless self-shrinking surfaces \(\Sigma^2\subset\mathbb R^N\), \(N>2\), and proves compactness for the low-entropy class \(\lambda(\Sigma)<2\). It then uses that compactness to establish the existence of entropy minimizers among non-flat self-shrinking surfaces, again in a geometric sense unrelated to 2D encoding trees [2108.03919].

A third, conceptually separate usage appears in the study of two-dimensional supercooled liquids. There entropy is estimated from the decay of recurring structural patterns with patch size through a higher-dimensional Grassberger–Procaccia construction. The central scaling law is
\[
C_m^{(d)}(\epsilon)\simeq c\,\epsilon^\nu
\exp\!\left[-d(m-1)s_m^{\mathrm{R\acute{e}nyi}}\right],
\]
so entropy is read off from the exponential decay rate of patch recurrence. This is a measurement framework for configurational entropy in 2D amorphous systems, not an optimization over graph partitions [2605.29945].

These literatures share the language of entropy minimization over two-dimensional objects, but they optimize or estimate fundamentally different quantities. A plausible implication is that “two-dimensional structural entropy minimization” is now a cross-disciplinary label rather than a single canonical formalism.

## 6. Interpretation, misconceptions, and limitations

A persistent misconception is that “two-dimensional” means planar geometry. In the dominant graph-partitioning literature, it usually means a height-2 encoding tree: root, clusters, leaves. This is explicit in social event detection, dialogue partitioning, and dynamic-graph work, and the image-segmentation paper goes out of its way to clarify that 2D SE there is “the structural entropy of a graph under a partition,” not entropy over pixel coordinates directly [2312.11891][2508.05023][2207.12653][2309.01899].

A second common misunderstanding is that 2D SE requires the number of clusters in advance. The opposite is central to several papers. HISEvent, SLED, MRGSEM-Sum, and the DiaASQ partitioner all present adaptive cluster-number selection as a primary advantage over \(K\)-means, NCut, spectral clustering, or graph-clustering pipelines that require a fixed cluster count [2312.11891][2309.01899][2507.23400][2508.05023].

The major technical limitation is that most practical optimizers are heuristic. HISEvent’s local subgraph processing, ADP-SEMEvent’s “optimal subgraphs,” MRGSEM-Sum’s batched merging, and DSEM’s greedy parallel merges are all approximate procedures rather than exact global solvers for the 2D SE objective. Incre-2dSE is especially explicit on this point: node shifting is a greedy local optimizer, may not converge, and requires an iteration cap [2407.18274][2507.23400][2508.05023][2207.12653].

Another limitation is dependence on graph construction. SLED reports that performance degrades when the spatial threshold \(r\) is too small, because graph construction underuses color-similarity structure; HISEvent’s entire design separates 1D SE for graph augmentation from 2D SE for graph partitioning, implicitly acknowledging that a poor graph can undermine even a strong partition objective [2309.01899][2312.11891].

Finally, some recent papers present formulas with noticeable typesetting corruption. The privacy-preserving event-detection paper states that its printed 2D structural entropy formula is malformed but recoverable from surrounding notation; the summarization paper’s fusion and merge equations are likewise described as corrupted; and the dialogue paper’s merge-score sign convention appears inconsistent between formula and prose. This suggests that, for implementation, the surrounding algorithmic definitions and the earlier Li–Pan structural-information formalism remain essential interpretive anchors [2407.18274][2507.23400][2508.05023].

Taken together, the literature presents two-dimensional structural entropy minimization as a flexible unsupervised partition principle whose main strengths are adaptive cluster discovery, compatibility with weighted relational graphs, and extensibility to privacy-preserving, dynamic, and multi-relational settings. Its main weaknesses are optimizer heuristics, graph-construction sensitivity, and occasional formal ambiguity in presentation.

Source: https://www.emergentmind.com/topics/two-dimensional-structural-entropy-minimization