Pairwise network comparison is essential for various applications, including neuroscience, disease research, and dynamic network analysis. While existing literature primarily focuses on comparing entire network structures, we address a vertex-wise comparison problem where two random networks share the same set of vertices but allow for structural variations in some vertices, enabling a more detailed and flexible analysis of network differences. In our framework, some vertices retain their latent positions between networks, while others undergo shifts. To identify the shifted and unshifted vertices and estimate their latent position shifts, we propose a method that first derives vertex embeddings in a low-rank Euclidean space for each network, then aligns these estimated vertex latent positions into a common space to resolve potential non-identifiability, and finally tests whether each vertex is shifted or not and estimates the vertex shifts. Our theoretical results establish the test statistic for the algorithms, guide parameter selection, and provide performance guarantees. Simulation studies and real data applications, including a case-control study in disease research and dynamic network analysis, demonstrate that the proposed algorithms are both computationally efficient and effective in extracting meaningful insights from network comparisons.

翻译：暂无翻译