Research map / Graph learning / Knowledge and temporal graphs
S3Former: Sequential, Structural, and Statistical Fusion for Continuous-Time Dynamic Graphs
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes S3Former, a fusion model for continuous-time dynamic graphs that combines sequential, structural, and statistical views of historical interactions to predict future links.
Problem
Predicting future links in continuous-time dynamic graphs with irregularly timed interactions, where existing memory-based and sequence-based methods organize history as node memories or serialized neighbor sequences and may miss local topology and historical relation patterns.
Approach
Fuses sequential (temporal order), structural (local topology), and statistical (historical relation pattern) information about interactions, rather than relying only on node memories or serialized neighbor sequences.
Closest papers by meaning
- A Deep Probabilistic Framework for Continuous Time Dynamic Graph Generation (AAAI 2025 · similarity 0.90)
- Future Link Prediction Without Memory or Aggregation (NeurIPS 2025 · similarity 0.89)
- TAMI: Taming Heterogeneity in Temporal Interactions for Temporal Graph Link Prediction (NeurIPS 2025 · similarity 0.88)
- State Space Models on Temporal Graphs: A First-Principles Study (NeurIPS 2024 · similarity 0.88)
- Analytical Construction on Geometric Architectures: Transitioning from Static to Temporal Link Prediction (ICML 2025 · similarity 0.88)
- DyG-Mamba: Continuous State Space Modeling on Dynamic Graphs (NeurIPS 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.