Get Rid of Isolation: A Continuous Multi-task Spatio-Temporal Learning Framework

TL;DR

CMuST tackles task isolation in urban spatiotemporal learning by bridging multiple urban datasets through a continuous multi-task framework. It introduces MSTI to capture cross-dimensional interactions (context-spatial, context-temporal) and RoAda to roll through tasks with task prompts and weight-behavior modeling for shared and task-specific patterns. Evaluated on NYC, SIP, and Chicago benchmarks with multiple tasks per city, CMuST outperforms single-task baselines and existing multi-task methods, especially under data-scarce and cold-start conditions, demonstrating improved generalization and data efficiency. This work advances collective urban intelligence and provides a scalable, adaptable platform for open urban systems, with code publicly available for reproduction.

Abstract

Spatiotemporal learning has become a pivotal technique to enable urban intelligence. Traditional spatiotemporal models mostly focus on a specific task by assuming a same distribution between training and testing sets. However, given that urban systems are usually dynamic, multi-sourced with imbalanced data distributions, current specific task-specific models fail to generalize to new urban conditions and adapt to new domains without explicitly modeling interdependencies across various dimensions and types of urban data. To this end, we argue that there is an essential to propose a Continuous Multi-task Spatio-Temporal learning framework (CMuST) to empower collective urban intelligence, which reforms the urban spatiotemporal learning from single-domain to cooperatively multi-dimensional and multi-task learning. Specifically, CMuST proposes a new multi-dimensional spatiotemporal interaction network (MSTI) to allow cross-interactions between context and main observations as well as self-interactions within spatial and temporal aspects to be exposed, which is also the core for capturing task-level commonality and personalization. To ensure continuous task learning, a novel Rolling Adaptation training scheme (RoAda) is devised, which not only preserves task uniqueness by constructing data summarization-driven task prompts, but also harnesses correlated patterns among tasks by iterative model behavior modeling. We further establish a benchmark of three cities for multi-task spatiotemporal learning, and empirically demonstrate the superiority of CMuST via extensive evaluations on these datasets. The impressive improvements on both few-shot streaming data and new domain tasks against existing SOAT methods are achieved. Code is available at https://github.com/DILab-USTCSZ/CMuST.

Figures (6)

• Figure 1: Illustration of evolution on data patterns and learning tasks.
• Figure 2: Framework overview of CMuST.
• Figure 3: Ablation studies of CMuST on Chicago.
• Figure 4: Case studies of our proposed CMuST on NYC.
• Figure 5: This figure illustrates the process of Data Representation and Integration within the CMuST framework. $T$ and $N$ denote the length of the time series and the number of geographic nodes, respectively. The diagram on the right details the encoding of information across different dimensions, with the bottom right corner aggregating the results of all encodings into a unified multidimensional representation.