We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Statistics > Machine Learning

Title: DIGRAC: Digraph Clustering Based on Flow Imbalance

Abstract: Node clustering is a powerful tool in the analysis of networks. We introduce a graph neural network framework to obtain node embeddings for directed networks in a self-supervised manner, including a novel probabilistic imbalance loss, which can be used for network clustering. Here, we propose directed flow imbalance measures, which are tightly related to directionality, to reveal clusters in the network even when there is no density difference between clusters. In contrast to standard approaches in the literature, in this paper, directionality is not treated as a nuisance, but rather contains the main signal. DIGRAC optimizes directed flow imbalance for clustering without requiring label supervision, unlike existing graph neural network methods, and can naturally incorporate node features, unlike existing spectral methods. Extensive experimental results on synthetic data, in the form of directed stochastic block models, and real-world data at different scales, demonstrate that our method, based on flow imbalance, attains state-of-the-art results on directed graph clustering when compared against 10 state-of-the-art methods from the literature, for a wide range of noise and sparsity levels, graph structures and topologies, and even outperforms supervised methods.
Comments: 40 pages (9.5 pages for main text)
Subjects: Machine Learning (stat.ML); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR); Machine Learning (cs.LG)
Cite as: arXiv:2106.05194 [stat.ML]
  (or arXiv:2106.05194v7 [stat.ML] for this version)

Submission history

From: Yixuan He [view email]
[v1] Wed, 9 Jun 2021 16:33:13 GMT (1141kb,D)
[v2] Wed, 6 Oct 2021 14:46:53 GMT (6111kb,D)
[v3] Sun, 10 Oct 2021 03:00:32 GMT (6123kb,D)
[v4] Wed, 13 Oct 2021 08:21:28 GMT (6045kb,D)
[v5] Mon, 31 Jan 2022 08:30:52 GMT (14153kb,D)
[v6] Mon, 23 May 2022 19:42:02 GMT (14413kb,D)
[v7] Sun, 18 Sep 2022 20:38:09 GMT (14697kb,D)

Link back to: arXiv, form interface, contact.