ComS2T: A Complementary Spatiotemporal Learning System for Data-Adaptive Model Evolution
- 5 June 2025
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 47 (10) , 8506-8523
Abstract
Spatiotemporal (ST) learning has become a crucial technique to enable smart cities and sustainable urban development. Current ST learning models capture the heterogeneity via various spatial convolution and temporal evolution blocks. However, rapid urbanization leads to fluctuating distributions in urban data and city structures, resulting in existing methods suffering generalization and data adaptation issues. Despite efforts, existing methods fail to deal with newly arrived observations, and the limitation of those methods with generalization capacity lies in the repeated training that leads to inconvenience, inefficiency and resource waste. Motivated by complementary learning in neuroscience, we introduce a prompt-based complementary spatiotemporal learning termed ComS2T, to empower the evolution of models for data adaptation. We first disentangle the neural architecture into two disjoint structures, a stable neocortex for consolidating historical memory, and a dynamic hippocampus for new knowledge update. Then we train the dynamic spatial and temporal prompts by characterizing distribution of main observations to enable prompts adaptive to new data. This data-adaptive prompt mechanism, combined with a two-stage training process, facilitates fine-tuning of the neural architecture conditioned on prompts, thereby enabling efficient adaptation during testing. Extensive experiments validate the efficacy of ComS2T in adapting various spatiotemporal out-of-distribution scenarios while maintaining effective inferences. The code is available on https://github.com/hqh0728/ComS2T.Keywords
All Related Versions
Funding Information
- National Natural Science Foundation of China (12227901, 62402414)
- Natural Science Foundation of Jiangsu Province (BK.20240460)
- State Key Laboratory of Resources and Environmental Information System
- Guangzhou Municipal Science and Technology (2023A03J0011)
This publication has 59 references indexed in Scilit:
- Deciphering urban traffic impacts on air quality by deep learning and emission inventoryJournal of Environmental Sciences, 2022
- Revisiting Convolutional Neural Networks for Citywide Crowd Flow Analytics2021
- STConvS2S: Spatiotemporal Convolutional Sequence to Sequence Network for weather forecastingNeurocomputing, 2021
- Clinical applications of continual learning machine learningThe Lancet Digital Health, 2020
- Deep learning tools for the measurement of animal behavior in neuroscienceCurrent Opinion in Neurobiology, 2020
- Biologically plausible learning in neural networks with modulatory feedbackNeural Networks, 2017
- What Learning Systems do Intelligent Agents Need? Complementary Learning Systems Theory UpdatedTrends in Cognitive Sciences, 2016
- Incorporating rapid neocortical learning of new schema-consistent information into complementary learning systems theory.Journal of Experimental Psychology: General, 2013
- Spatial and Temporal Dependence in House Price PredictionThe Journal of Real Estate Finance and Economics, 2012
- Overview of Artificial Neural Networks2008