Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

UniCom: Towards a Unified and Cohesiveness-aware Framework for Community Search and Detection

Yifan Zhu, Hanchen Wang, Wenjie Zhang, Alexander Zhou, Ying Zhang

TL;DR

UniCom introduces a cohesiveness-aware, multi-domain framework that unifies community search and detection. It leverages Domain-aware Specialization and Universal Graph Learning with cohesive subgraph prompts, conductance-based local subgraphs, and lightweight domain adaptation prompts to transfer knowledge across domains while freezing the backbone. A multi-domain fusion strategy and task-specific experts enable robust CS/CD performance under scarce supervision, validated across 16 datasets and 22 baselines with strong efficiency. The work provides theoretical insights and practical design choices that position UniCom as a foundation model for subgraph-level tasks across domains.

Abstract

Searching and detecting communities in real-world graphs underpins a wide range of applications. Despite the success achieved, current learning-based solutions regard community search, i.e., locating the best community for a given query, and community detection, i.e., partitioning the whole graph, as separate problems, necessitating task- and dataset-specific retraining. Such a strategy limits the applicability and generalization ability of the existing models. Additionally, these methods rely heavily on information from the target dataset, leading to suboptimal performance when supervision is limited or unavailable. To mitigate this limitation, we propose UniCom, a unified framework to solve both community search and detection tasks through knowledge transfer across multiple domains, thus alleviating the limitations of single-dataset learning. UniCom centers on a Domain-aware Specialization (DAS) procedure that adapts on the fly to unseen graphs or tasks, eliminating costly retraining while maintaining framework compactness with a lightweight prompt-based paradigm. This is empowered by a Universal Graph Learning (UGL) backbone, which distills transferable semantic and topological knowledge from multiple source domains via comprehensive pre-training. Both DAS and UGL are informed by local neighborhood signals and cohesive subgraph structures, providing consistent guidance throughout the framework. Extensive experiments on both tasks across 16 benchmark datasets and 22 baselines have been conducted to ensure a comprehensive and fair evaluation. UniCom consistently outperforms all state-of-the-art baselines across all tasks under settings with scarce or no supervision, while maintaining runtime efficiency.

UniCom: Towards a Unified and Cohesiveness-aware Framework for Community Search and Detection

TL;DR

UniCom introduces a cohesiveness-aware, multi-domain framework that unifies community search and detection. It leverages Domain-aware Specialization and Universal Graph Learning with cohesive subgraph prompts, conductance-based local subgraphs, and lightweight domain adaptation prompts to transfer knowledge across domains while freezing the backbone. A multi-domain fusion strategy and task-specific experts enable robust CS/CD performance under scarce supervision, validated across 16 datasets and 22 baselines with strong efficiency. The work provides theoretical insights and practical design choices that position UniCom as a foundation model for subgraph-level tasks across domains.

Abstract

Searching and detecting communities in real-world graphs underpins a wide range of applications. Despite the success achieved, current learning-based solutions regard community search, i.e., locating the best community for a given query, and community detection, i.e., partitioning the whole graph, as separate problems, necessitating task- and dataset-specific retraining. Such a strategy limits the applicability and generalization ability of the existing models. Additionally, these methods rely heavily on information from the target dataset, leading to suboptimal performance when supervision is limited or unavailable. To mitigate this limitation, we propose UniCom, a unified framework to solve both community search and detection tasks through knowledge transfer across multiple domains, thus alleviating the limitations of single-dataset learning. UniCom centers on a Domain-aware Specialization (DAS) procedure that adapts on the fly to unseen graphs or tasks, eliminating costly retraining while maintaining framework compactness with a lightweight prompt-based paradigm. This is empowered by a Universal Graph Learning (UGL) backbone, which distills transferable semantic and topological knowledge from multiple source domains via comprehensive pre-training. Both DAS and UGL are informed by local neighborhood signals and cohesive subgraph structures, providing consistent guidance throughout the framework. Extensive experiments on both tasks across 16 benchmark datasets and 22 baselines have been conducted to ensure a comprehensive and fair evaluation. UniCom consistently outperforms all state-of-the-art baselines across all tasks under settings with scarce or no supervision, while maintaining runtime efficiency.

Paper Structure

This paper contains 44 sections, 1 theorem, 29 equations, 6 figures, 8 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Works
  3. Community Search and Detection
  4. Graph Transfer Learning
  5. Preliminaries
  6. Problem Statement
  7. Methodology
  8. Domain-aware Specialization
  9. Graph Augmentation and Alignment
  10. Local Subgraph Construction
  11. Cohesive Subgraph Prompt
  12. Domain Adaptation Prompt
  13. Feature Alignment
  14. Graph Transformer Architecture
  15. Community Task Expert
  16. ...and 29 more sections

Key Result

Lemma 1

For the multi-domain knowledge fusion, the CS classification risk never exceeds that of each individual prediction. The same monotonicity property applies to the DCD joint distortion and the OCD self-supervised loss.

Figures (6)

  • Figure 1: Overview of core tasks handled by UniCom. (a) Community Search (CS), (b) Disjoint Community Detection (DCD), and (c) Overlapping Community Detection (OCD).
  • Figure 2: Dataset Pre-processing.
  • Figure 3: Overview of UniCom. 1. In Universal Graph Learning (UGL), we pre-train multiple models in parallel using datasets from different domains. 2. For Domain-aware Specialization (DAS), the pre-trained models are frozen, while anchor nodes selected from pre-training datasets are transferred. 3. Finally, outputs from multiple models are fused for the downstream task.
  • Figure 4: Total Online Searching Time of Community Search (100 Queries).
  • Figure 5: Training Time of Community Search and Detection.
  • ...and 1 more figures

Theorems & Definitions (5)

  • definition 1: Community Search ICS-GNNSMN
  • definition 2: Community Detection cd_definitionUCoDe
  • definition 3: Conductance conductance1conductance2
  • Example 1
  • Lemma 1: Risk Reduction with Fusion