Saved in:
Bibliographic Details
Main Authors: Dayag, Elisha, Van Tran, Nhat Thanh, Xin, Jack
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2508.20206
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917481338109952
author Dayag, Elisha
Van Tran, Nhat Thanh
Xin, Jack
author_facet Dayag, Elisha
Van Tran, Nhat Thanh
Xin, Jack
contents Transformer-based models are at the forefront in long time-series forecasting (LTSF). While in many cases, these models are able to achieve state of the art results, they suffer from a bias toward low-frequencies in the data and high computational and memory requirements. Recent work has established that learnable frequency filters can be an integral part of a deep forecasting model by enhancing the model's spectral utilization. These works choose to use a multilayer perceptron to process their filtered signals and thus do not solve the issues found with transformer-based models. In this paper, we establish that adding a filter to the beginning of transformer-based models enhances their performance in long time-series forecasting. We add learnable filters, which only add an additional $\approx 1000$ parameters to several transformer-based models and observe in multiple instances 5-10 \% relative improvement in forecasting performance. Additionally, we find that with filters added, we are able to decrease the embedding dimension of our models, resulting in transformer-based architectures that are both smaller and more effective than their non-filtering base models. We also conduct synthetic experiments to analyze how the filters enable Transformer-based models to better utilize the full spectrum for forecasting.
format Preprint
id arxiv_https___arxiv_org_abs_2508_20206
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Filter then Attend: Improving attention-based Time Series Forecasting with Spectral Filtering
Dayag, Elisha
Van Tran, Nhat Thanh
Xin, Jack
Machine Learning
Artificial Intelligence
Transformer-based models are at the forefront in long time-series forecasting (LTSF). While in many cases, these models are able to achieve state of the art results, they suffer from a bias toward low-frequencies in the data and high computational and memory requirements. Recent work has established that learnable frequency filters can be an integral part of a deep forecasting model by enhancing the model's spectral utilization. These works choose to use a multilayer perceptron to process their filtered signals and thus do not solve the issues found with transformer-based models. In this paper, we establish that adding a filter to the beginning of transformer-based models enhances their performance in long time-series forecasting. We add learnable filters, which only add an additional $\approx 1000$ parameters to several transformer-based models and observe in multiple instances 5-10 \% relative improvement in forecasting performance. Additionally, we find that with filters added, we are able to decrease the embedding dimension of our models, resulting in transformer-based architectures that are both smaller and more effective than their non-filtering base models. We also conduct synthetic experiments to analyze how the filters enable Transformer-based models to better utilize the full spectrum for forecasting.
title Filter then Attend: Improving attention-based Time Series Forecasting with Spectral Filtering
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2508.20206