XLinear: Frequency-Enhanced MLP with CrossFilter for Robust Long-Range Forecasting

TL;DR

XLinear is a novel MLP-based time series forecasting model that effectively captures long-range dependencies using frequency-domain operations and a CrossFilter Block, achieving state-of-the-art results while maintaining robustness.

Contribution

The paper introduces XLinear, combining frequency-enhanced attention and a CrossFilter Block to improve long-range forecasting with an MLP-based model.

Findings

XLinear outperforms existing MLP-based forecasters on benchmark datasets.

The model effectively captures long-term dependencies.

It maintains high robustness to noise.

Abstract

Time series forecasters are widely used across various domains. Among them, MLP (multi-layer perceptron)-based forecasters have been proven to be more robust to noise compared to Transformer-based forecasters. However, MLP struggles to capture complex features, resulting in limitations on capturing long-range dependencies. To address this challenge, we propose XLinear, an MLP-based forecaster for long-range forecasting. Firstly, we decompose the time series into trend and seasonal components. For the trend component which contains long-range characteristics, we design Enhanced Frequency Attention (EFA) to capture long-term dependencies by leveraging frequency-domain operations. Additionally, a CrossFilter Block is proposed for the seasonal component to maintain the model's robustness to noise, avoiding the problems of low robustness often caused by attention mechanisms. Experimental results demonstrate that XLinear achieves state-of-the-art performance on test datasets. While keeping the lightweight architecture and high robustness of MLP-based models, our forecaster outperforms other MLP-based forecasters in capturing long-range dependencies.