A new review paper explores the application of optimal transport methods for comparing networks, particularly in machine learning contexts. The paper details three primary distances: Wasserstein, Gromov-Wasserstein, and Bures-Wasserstein, and examines their properties for undirected, unweighted graphs. It also discusses how these distances can reveal node influences and provides methods for bounding distances using Laplacian spectra. The review concludes by evaluating these optimal transport distances for clustering and anomaly detection tasks using both synthetic and real-world network data. AI
IMPACT This review could advance machine learning applications in network analysis by providing a unified framework for comparing complex network structures.
RANK_REASON The item is a review paper published on arXiv detailing a specific methodology (optimal transport) for network comparison with machine learning applications. [lever_c_demoted from research: ic=1 ai=1.0]
- anomaly detection
- arXiv
- Bures-Wasserstein distance
- clustering algorithm
- Graph Metrics of Structural Brain Networks in Individuals with Schizophrenia and Healthy Controls: Group Differences, Relationships with Intelligence, and Genetics
- Gromov-Wasserstein distance
- Laplacian spectra of, and random walks on, complex networks: are scale-free architectures really important?
- Network Comparison of Inflammation in Colorectal Cancer and Alzheimer's Disease
- Network science (Cambridge University Press)
- optimal transport
- Wasserstein metric
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
