TIM: Interpretable Modelling of Complex Temporal Interactions in Multivariate Networks
Keywords: Time Series, Deep Learning, Decomposition
TL;DR: An Interpretable Modelling of Frequency Domain and Temporal Interactions in Multivariate Networks for LSTF
Abstract: Multivariate time series forecasting is crucial across various fields and essential for addressing numerous real-world challenges. However, existing forecasting methods have significant limitations: while Transformer models are effective, they are constrained by high computational costs and declining performance in long-term forecasting; MLP models struggle to capture complex multivariate interactions. These issues hinder the models' ability to accurately decompose seasonality and trends. To tackle these problems, we propose a new method called TIM. Through a cross-layer architecture, TIM decomposes time series predictions into temporal features, multivariate interaction features, and residual components. Our all-MLP model integrates global features with complex multivariate dynamics. By introducing a linear self-attention mechanism across variables and time steps, TIM enhances the learning of feature interactions and accurately captures temporal transitions between domains. This innovative design leverages linear attention mechanisms and cross-layer architecture to more effectively model temporal features and multivariate interactions. It surpasses traditional Transformer-based methods by improving predictive accuracy while maintaining linear computational complexity. Experimental results demonstrate that TIM outperforms existing state-of-the-art methods while ensuring computational efficiency.
Primary Area: learning on time series and dynamical systems
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 2432
Loading