Multi-Level Spatial Embedding Sharing for Enhanced Online Trajectory-User Linking

TMLR Paper6881 Authors

Abstract: Trajectory-User Linking (TUL) is a critical task in mobility applications that links unlabeled spatial trajectories to the users or entities that generated them. In these applications, data often arrives as a continuous stream and may experience distributional shifts over time. While adapting TUL models via online learning could address these challenges, this approach remains unexplored in current research. Our work bridges this gap by conducting comprehensive evaluations of common TUL techniques in an online learning context. To improve the performance of existing TUL techniques in this setting, we further introduce \underline{M}ult\underline{i}-\underline{L}evel Spatial \underline{E}mbedding \underline{S}haring (MiLES), an embedding approach that adapts multi-level spatial representation to the online TUL setting. MiLES operates by partially sharing embeddings for locations within neighborhoods of multiple size levels. This design enables generalization of knowledge within neighborhoods, while maintaining fine-grained discrimination through more location-specific representations. MiLES also significantly reduces the number of embedding parameters leading to lower memory usage and more computationally efficient model updates. We further incorporate learnable weighting parameters for each embedding level, allowing the model to learn the influence of different levels during training. Our experimental results on several real-world datasets show that integrating MiLES into state-of-the-art TUL models significantly improves their performance in online learning scenarios, yielding relative gains in top-1 accuracy of up to 24\%. To demonstrate its general applicability, we also evaluate MiLES on the task of destination prediction, where it also provides consistent performance improvements, confirming its value as a domain-general embedding technique. Our code is available at \url{https://anonymous.4open.science/r/MiLES-3D20}.

Submission Type: Long submission (more than 12 pages of main content)

Changes Since Last Submission:

**Privacy motivation**
Clarified the privacy argument in the introduction to specify the concrete mechanism: online learning allows trajectory data to be discarded immediately after each update, reducing breach risk, rather than claiming a general privacy advantage.
Revised the broader impact statement to distinguish between the capability risk (a function of model accuracy, independent of training paradigm) and the data retention advantage specific to online learning.

**Positioning against periodic retraining**
No additional changes to the paper, as the revised text already acknowledges periodic retraining as a practical alternative and lists its investigation as future work. We would like to clarify that the paper positions online learning as a viable and previously unstudied framework for TUL, rather than claiming superiority over other paradigms like periodic retraining.

Assigned Action Editor: ~Yan_Liu1

Submission Number: 6881