BEAT: Balanced Frequency Adaptive Tuning for Long-Term Time-Series Forecasting

TL;DR

BEAT introduces a dynamic frequency-aware training framework for long-term time-series forecasting, balancing learning across frequencies to improve accuracy and prevent overfitting.

Contribution

This paper presents BEAT, a novel adaptive tuning method that monitors and adjusts frequency-specific training to enhance long-term forecasting performance.

Findings

BEAT outperforms existing methods on seven real-world datasets.

Dynamic frequency balancing improves convergence and accuracy.

The approach effectively prevents overfitting of high-frequency components.

Abstract

Time-series forecasting is crucial for numerous real-world applications including weather prediction and financial market modeling. While temporal-domain methods remain prevalent, frequency-domain approaches can effectively capture multi-scale periodic patterns, reduce sequence dependencies, and naturally denoise signals. However, existing approaches typically train model components for all frequencies under a unified training objective, often leading to mismatched learning speeds: high-frequency components converge faster and risk overfitting, while low-frequency components underfit due to insufficient training time. To deal with this challenge, we propose BEAT (Balanced frEquency Adaptive Tuning), a novel framework that dynamically monitors the training status for each frequency and adaptively adjusts their gradient updates. By recognizing convergence, overfitting, or underfitting for each frequency, BEAT dynamically reallocates learning priorities, moderating gradients for rapid learners and increasing those for slower ones, alleviating the tension between competing objectives across frequencies and synchronizing the overall learning process. Extensive experiments on seven real-world datasets demonstrate that BEAT consistently outperforms state-of-the-art approaches.