Network comparison plays a central role in characterizing structural differences and cross-network correlations in complex systems. In many real-world settings, however, interactions are inherently signed, with positive and negative links altering both connection semantics and structural organization. This challenges conventional comparison methods built upon unsigned assumptions, which are unable to adequately capture such heterogeneity. To address this limitation, we introduce a network comparison method based on Signed Communicability Embedding (SCE). SCE employs the matrix exponential of the signed adjacency matrix to capture the cumulative contributions of positive and negative walks across multiple scales. Network-level dissimilarity is then quantified through discrepancies in pairwise node distances within the resulting embedding space, thereby integrating topological structure and relation polarity into a unified measure. To further ensure consistency across networks, a spectral correction strategy is incorporated to mitigate scale-induced bias. Extensive experiments on diverse real-world signed networks show that SCE yields stable and discriminative performance under a variety of perturbation scenarios, particularly in capturing structural shifts induced by negative edge changes. Additional analyses based on null models and network clustering further show that SCE not only disentangles differences arising from topology and sign configurations but also organizes networks into distinct structural regimes, reflecting variations in connectivity density, local closure, and signed interaction heterogeneity. Overall, SCE provides a coherent and interpretable approach to network comparison in complex signed systems.
A statistical filter that benchmarks node-level bipartite clustering against degree-preserving randomizations to classify nodes as geometric (signal) or degree constrained noise is introduced, offering a simple, scalable way to disentangle structure from noise in bipartite networks.
L. Ramirez, Roya Aliakbarisani, M. Serrano et al.· Nature Communications· 0 citations
This study introduces a new ranking framework that integrates a quasi-Laplacian structural measure with a gravity-inspired aggregation process and demonstrates that the proposed framework consistently outperforms existing techniques in terms of accuracy, resolution, and computational simplicity.
Many real-world systems—ranging from protein structures to financial networks—are naturally represented as labeled graphs, where both topology and node attributes carry critical information. A fundamental question in analyzing such data is whether two graphs (or subgraphs) exhibit statistical dependence, which may indicate shared generative mechanisms or latent interactions. Unlike classical dependence testing, the graph setting introduces unique challenges: dependence can manifest through structural similarity, label correlation, or their interplay, potentially reinforcing or obscuring each other. We propose a novel and practical framework for dependence testing in labeled graphs via mutual information over a structure-weighted joint label distribution. This approach jointly captures topological and attribute-based signals while remaining robust to imperfect or noisy node alignments. We provide theoretical guarantees with explicit error bounds and validate our method on both synthetic and real-world datasets, including protein structures from the lipocalin family, and recurring motifs in the Cora citation network. Our results demonstrate that the proposed test is a statistically sound and an effective tool for uncovering nontrivial dependencies in graph data.
Nikolaos Papagiannis, Vasam Manjveekar Prabantu, A. Grama et al.· Proceedings of the 32nd ACM...· 0 citations
While signed social recommendation has shown great potential by modeling both trust and distrust relations, its effectiveness is often hindered by structural noise and data sparsity. In this work, we first identify a fundamental inconsistency across the structural, propagation, and semantic layers of existing models, which leads to biased representations learned from sparse or noisy datasets. Furthermore, we observe that most existing methods treat the observed graph as fixed, failing to bridge the gap between noisy topologies and reliable social semantics. To address these issues, we propose a unified framework named SSC-Loop that treats signed social recommendation as the maximization of structural consistency. SSC-Loop includes three dedicated modules: ESA-DA for structural consistency, a P/N/O propagation mechanism for propagation consistency, and a contrastive learning objective for semantic consistency. Experiments on Epinions demonstrate that SSC-Loop achieves strong performance on explicit signed social rating prediction, while auxiliary results on Slashdot under a derived link-existence setting further suggest its ability to exploit signed social structures. Source code is available at https://github.com/Refrainwww/SSC-Loop.
Polarization is common in social systems, where individuals tend to form cohesive groups that oppose each other. Signed networks, with positive edges representing agreement and negative edges representing disagreement, provide a natural model for studying such dynamics. The 2-Polarized-Communities problem (2pc) was recently introduced to detect a single pair of polarized communities by maximizing a Rayleigh quotient that balances intra-community agreement and inter-community disagreement. However, real signed networks usually host multiple, coexisting axes of conflict, often with communities that overlap. Existing extension of 2pc to multiple communities or find-and-remove heuristics, either rely on the restrictive assumption that every polarized community is in conflict with all the others, or enforce disjoint solutions–thus failing to capture the nuanced structures observed in real networks. In this paper, we introduce the Diverse top-k-pc problem, which is the first principled formulation of top-k polarized communities with controlled overlap. Our formulation extends the 2pc polarity objective by incorporating diversity terms directly into the denominator of the Rayleigh quotient, yielding a generalized objective that jointly promotes polarity and diversity. We design a greedy sequential algorithm that solves a generalized eigenvector problem at each step, efficiently discovering diverse polarized pairs. Experiments on both real-world and synthetic signed networks demonstrate that our approach identifies multiple meaningful and overlapping pairs of polarized communities, outperforming natural baselines while scaling to large graphs.
Francesco Gullo, Domenico Mandaglio, Andrea Tagarelli· Machine-mediated learning· 0 citations