Community Search: A Meta-Learning Approach

Abstract

Community Search (CS) is one of the fundamental graph analysis tasks, which is a building block of various real applications. Given any query nodes, CS aims to find cohesive subgraphs that query nodes belong to. Recently, a large number of CS algorithms are designed. These algorithms adopt pre-defined subgraph patterns to model the communities, which cannot find communities that do not have such pre-defined patterns in real-world graphs. Thereby, machine learning based approaches are proposed to capture flexible community structures by learning from community ground-truth in a data-driven fashion. However, existing approaches rely on sufficient training data to provide enough generalization for machine learning models.In this paper, we study ML-based approaches for community search, under the circumstance that the training data is scarce. To learn from small data, we extract prior knowledge which is shared across different graphs as CS tasks in advance. Subsequent small training data from a new CS task are combined with the learned prior knowledge to help the model well adapt to that specific task. A novel meta-learning based framework, called CGNP, is designed and implemented to fulfill this learning procedure. A meta CGNP model is a task-common node embedding function for clustering by nature, learned by metric-based learning. To the best of our knowledge, CGNP is the first meta model solution for CS. We compare CGNP with traditional CS algorithms, e.g., CTC, ATC, ACQ, and ML baselines on real graph datasets with groundtruth. Our experiments show that CGNP outperforms the native graph algorithms and ML baselines 147% and 113% on F1-score by average.

MetaGNN-Based Medical Records Unstructured Specialized Vocabulary Few-Shot Representation Learning

“…Comparison with current state-of-theart related working model results in dataset Reddit. These related works include 2019-2022 related graph representation learning models such as Meta SGC [40], GCN [41], SemiGNN [42], CGNP [43], etc. Based on these models, we have modified the dataset to meet the knowledge representation of meta-learning.…”

“…It can be seen from table 5 that the metalearning classifier based on graph knowledge representation can achieve 20.71% accuracy on a 5-way 3-shot compared with other classifiers. 9.47% 15.89% GraphSAGE-Pool [39] 9.31% 15.36% SGC [40] 9.80% 16.98% GCN [41] 9.87% 17.17% SemiGNN [42] 13.04% 15.07% CGNP [43] 15.38% 16.04% OURS 14.22% 20.71%…”