Research map / Label-efficient and robust learning / Recognition and long-tail
TrajAgg: Dual-Scale Feature Aggregation with Hybrid Training for Trajectory Similarity Computation in Free Space
Xiao Zhang, Xingyu Zhao, Yuan Cao, Bin Wang, Guiyuan Jiang, Yanwei Yu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add pairwise learning to embedding-only training (hybrid training) (ablation) | TrajAgg · Porto (Hausdorff) | HR@1: 60.26% -> 61.87% | Improved |
| add embedding learning to pairwise-only training (hybrid training) (ablation) | TrajAgg · Porto (DTW) | HR@1: 47.39% -> 47.59% | Mixed |
| add pairwise learning to embedding-only training (ablation) | TrajAgg · Porto (Discrete Frechet) | HR@1: 64.23% -> 65.74% | Improved |
| hybrid training instead of pairwise only (ablation) | TrajAgg · T-Driver (DTW) | HR@1: 14.01% -> 23.34% | Improved |
| hybrid training instead of embedding only (ablation) | TrajAgg · Geolife (Hausdorff) | HR@1: 45.01% -> 44.47% | Hurt |
| add the aggregation transformer (ablation) | TrajAgg · Porto (4 distances) | HR@1: approximately 31.67% average degradation when removed | Improved |
| use aggregation attention instead of vanilla self-attention (ablation) | TrajAgg · Porto (Hausdorff) | HR@1: 15.85% degradation when replaced | Improved |
| use aggregation attention instead of vanilla self-attention (ablation) | TrajAgg · Porto (Discrete Frechet) | HR@1: 8.82% degradation when replaced | Improved |
| add GPS features alongside grid features (ablation) | TrajAgg · Porto | HR@1 | Improved |
| add grid features alongside GPS features (ablation) | TrajAgg · Porto | HR@1 | Improved |
| use larger grid cell size (ablation) | TrajAgg · Porto | HR@1 | Hurt |
| increase weight mu on grid embedding (ablation) | TrajAgg · Porto | HR@1 | Hurt |
Closest papers by meaning
- UniTraj: Learning a Universal Trajectory Foundation Model from Billion-Scale Worldwide Traces (NeurIPS 2025 · similarity 0.86)
- TrajMamba: An Efficient and Semantic-rich Vehicle Trajectory Pre-training Model (NeurIPS 2025 · similarity 0.85)
- GTR: A General, Multi-View, and Dynamic Framework for Trajectory Representation Learning (ICML 2025 · similarity 0.85)
- TrajAgent: An LLM-Agent Framework for Trajectory Modeling via Large-and-Small Model Collaboration (NeurIPS 2025 · similarity 0.85)
- KGTS: Contrastive Trajectory Similarity Learning over Prompt Knowledge Graph Embedding (AAAI 2024 · similarity 0.84)
- Pedestrian Trajectory Prediction with Missing Data: Datasets, Imputation, and Benchmarking (NeurIPS 2024 · similarity 0.84)
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.
