SWGA: A Distributed Hyperparameter Search Method for Time Series Prediction Models
Keywords: Machine Learning, Deep Learning, Time Series Prediction, Hyperparameter Search, Genetic Algorithms
Abstract: We propose a distributed hyperparameter search method for time series prediction models named SWGA (Sliding Window Genetic Algorithm). Compared to current genetic algorithms for hyperparameter search, our method has three major advantages: (i) It adopts a configurable sliding window mechanism to effectively combat overfitting from distribution shifts inherent in time series data. (ii) It introduces a warm-up stage using Bayesian optimization-based methods to generate a good initial population. (iii) It supports distributed hyperparameter search across multi-node computing clusters, enhancing both scalability and efficiency. To demonstrate SWGA's efficacy, we conduct hyperparameter search experiments on time series datasets from various domains. The experiment results show that our method consistently finds a hyperparameter configuration that achieves better performance on out-of-sample time series data compared to the traditional genetic algorithm. On average, it reduces the out-of-sample loss by about 56.1%.
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: 12331
Loading