Guardado en:
| Autores principales: | , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | https://arxiv.org/abs/2505.08320 |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866918109006266368 |
|---|---|
| author | Choi, Yoonhyuk Choi, Jiho Kim, Chong-Kwon |
| author_facet | Choi, Yoonhyuk Choi, Jiho Kim, Chong-Kwon |
| contents | Recent Graph Neural Networks (GNNs) combine spectral-spatial architectures for enhanced representation learning. However, limited attention has been paid to certified robustness, particularly regarding training strategies and underlying rationale. In this paper, we explicitly specialize each branch: the spectral network is trained to withstand l0 edge flips and capture homophilic structures, while the spatial part is designed to resist linf feature perturbations and heterophilic patterns. A context-aware gating network adaptively fuses the two representations, dynamically routing each node's prediction to the more reliable branch. This specialized adversarial training scheme uses branch-specific inner maximization (structure vs feature attacks) and a unified alignment objective. We provide theoretical guarantees: (i) expressivity of the gating mechanism beyond 1-WL, (ii) spectral-spatial frequency bias, and (iii) certified robustness with trade-off. Empirically, SpecSphere attains state-of-the-art node classification accuracy and offers tighter certified robustness on real-world benchmarks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_08320 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Adaptive Branch Specialization in Spectral-Spatial Graph Neural Networks for Certified Robustness Choi, Yoonhyuk Choi, Jiho Kim, Chong-Kwon Machine Learning Recent Graph Neural Networks (GNNs) combine spectral-spatial architectures for enhanced representation learning. However, limited attention has been paid to certified robustness, particularly regarding training strategies and underlying rationale. In this paper, we explicitly specialize each branch: the spectral network is trained to withstand l0 edge flips and capture homophilic structures, while the spatial part is designed to resist linf feature perturbations and heterophilic patterns. A context-aware gating network adaptively fuses the two representations, dynamically routing each node's prediction to the more reliable branch. This specialized adversarial training scheme uses branch-specific inner maximization (structure vs feature attacks) and a unified alignment objective. We provide theoretical guarantees: (i) expressivity of the gating mechanism beyond 1-WL, (ii) spectral-spatial frequency bias, and (iii) certified robustness with trade-off. Empirically, SpecSphere attains state-of-the-art node classification accuracy and offers tighter certified robustness on real-world benchmarks. |
| title | Adaptive Branch Specialization in Spectral-Spatial Graph Neural Networks for Certified Robustness |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2505.08320 |