MRFNet: Multi-Receptive Field Network for Multivariate Time-Series Prediction
Keywords: time series forecasting
Abstract: Time series forecasting is a critical topic in machine learning. Although existing deep learning methods have demonstrated outstanding performance and currently dominate this field, the latest state-of-the-art (SOTA) models are increasingly encountering the same limitations: the blockneck of performance. We believe this convergence is due to these models being based on the same mathematical foundations. To address this issue, we draw inspiration from the universal approximation theorem (UAT) and show that most commonly used deep learning models for time series forecasting are specific implementations of UAT. Based on UAT theory and the characteristics of time series data, we propose a new forecasting model called the Multi-Receptive Field Network (MRFNet). This architecture integrates linear, sparse matrix, convolutional, and Fourier transform modules, resulting in an interpretable model with multiple receptive fields that can capture both global and local information. The MRFNet model has been tested extensively on several popular time series forecasting datasets and has achieved superior results.
Supplementary Material: pdf
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: 4170
Loading