Nov 2026· Proceedings of the 34th ACM International Conference on Multimedia· pp. 15397-15406· 0 citations· 24 references
TL;DR
Qualitative analyses suggest that the learned representations exhibit more compact intra-class structure and clearer inter-class separation, supporting ECHO as an effective approach to semi-supervised multi-view node classification.
Abstract
Learning discriminative representations from multi-view data remains challenging because different views exhibit heterogeneous structural patterns across scales. Many existing approaches model local and global relations in a single geometric space, which may limit their ability to jointly characterize fine-grained neighborhoods and non-uniform global organization. To address this limitation, we propose the Euclidean-hyperbolic correlational network, termed ECHO, a cross-scale collaborative learning framework. ECHO comprises a Poincaré-informed global perception module that performs non-local structural routing through a compact set of supernodes and a Euclidean local perception module for neighborhood aggregation. Within each view, a learnable scalar gate balances neighborhood information from the Euclidean branch with non-local structural context from the Poincaré-informed branch. Consensus-guided attention subsequently refines the resulting view-specific representations, with the mean representation across views serving as a consensus query to selectively integrate complementary information while preserving view-specific contributions. Across eight benchmark datasets, ECHO outperforms the compared methods in semi-supervised node classification in terms of both classification accuracy and Macro-F1. Qualitative analyses suggest that the learned representations exhibit more compact intra-class structure and clearer inter-class separation, supporting ECHO as an effective approach to semi-supervised multi-view node classification.
The quality of unsupervised manifold learning relies fundamentally on the underlying neighborhood graph. Standard linear methods utilizing a single affinity matrix often fail to capture the multiscale complexity of high-dimensional data. While multi-view algorithms integrate complementary information, additional views...
Qi-Rui Huang, Xuan S. Yang, Yang Zhao et al.· Proceedings of the 35th ACM...· 0 citations
Multi-view learning seeks to learn more comprehensive representations by exploiting the complementarity and consistency across diverse modalities or views. However, existing multi-view fusion strategies treat intra- and inter-view fusion as independent stages, without simultaneously considering the evolution within vie...
Yong-Quan Shi, Wei-Jun Huang, Yue-Yang Pi et al.· Proceedings of the 34th ACM...· 0 citations
Visualizing and interpreting the dependence structure in relational data often benefits from geometric representations via low-dimensional Euclidean embeddings. However, existing approaches are typically restricted to specific response types, rely on burdensome high-dimensional nonconvex optimization, or lack uniform t...
Across five benchmarks spanning sparse, dense, featureless, featureless, and attributed regimes, Concorde establishes state-of-the-art early retrieval precision, nearly doubling P@10% over leading baselines in sparse settings.
Cross-view geo-localization (CVGL) aims to retrieve images depicting the same geographic target across heterogeneous viewpoints. Drone–satellite matching remains challenging because holistic descriptors may dilute localized geographic cues, predefined partitions are sensitive to viewpoint- and altitude-induced spatial...
Fa-Wei Ge, Xin-Ao Wang, Lei-Po Liu et al.· IEEE Transactions on Geoscie...· 0 citations
This paper proposes a novel framework termed Few-shot Anchor-guided Multi-view Clustering with Pearson Correlation (FAMC-PC), which establishes a tuning-free design, substantially reducing the computational burden and labor costs associated with manual hyper-parameter tuning.
Song-Tao Li, Yi-Peng Wang, Yi-Tong Fan et al.· IEEE Transactions on Image P...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.