Non-Backtracking Matrix for Node Clustering
Research Paper#Graph Theory, Node Clustering, Machine Learning🔬 Research|Analyzed: Jan 3, 2026 09:29•
Published: Dec 30, 2025 19:38
•1 min read
•ArXivAnalysis
This paper explores the use of the non-backtracking transition probability matrix for node clustering in graphs. It leverages the relationship between the eigenvalues of this matrix and the non-backtracking Laplacian, developing techniques like "inflation-deflation" to cluster nodes. The work is relevant to clustering problems arising from sparse stochastic block models.
Key Takeaways
- •Investigates the use of the non-backtracking transition probability matrix for node clustering.
- •Explores the relationship between the eigenvalues of the non-backtracking matrix and the non-backtracking Laplacian.
- •Develops "inflation-deflation" techniques for clustering.
- •Applicable to clustering problems from sparse stochastic block models.
Reference / Citation
View Original"The paper focuses on the real eigenvalues of the non-backtracking matrix and their relation to the non-backtracking Laplacian for node clustering."