MSMixer: Learned Multi-Scale Temporal Mixing with Complementary Linear Shortcut for Long-Term Time Series Forecasting

TL;DR

MSMixer is a lightweight, multi-scale MLP model for long-term time series forecasting that effectively captures patterns at multiple temporal resolutions, outperforming several baselines.

Contribution

The paper introduces MSMixer, a novel multi-scale MLP architecture with dynamic weighting and a trend-seasonality shortcut, enabling efficient long-term forecasting.

Findings

MSMixer achieves the lowest average MSE among lightweight models on ETT benchmarks.

MSMixer outperforms DLinear and NLinear in 12 of 16 configurations.

MSMixer uses 5x fewer parameters than Transformer-based baselines while maintaining high accuracy.

Abstract

Long-term time series forecasting requires models that simultaneously capture rapid oscillations, medium-range periodicities, and slowly evolving macro-trends from a fixed look-back window. Existing lightweight MLP-based models typically operate on a single temporal resolution, limiting their ability to explicitly model patterns at multiple scales. We propose MSMixer, a channel-independent multi-scale MLP architecture that addresses this limitation through three complementary innovations: (i) three parallel scale branches at down-sample factors {1x, 4x, 16x} with independent MLP blocks, (ii) a learnable softmax gate that dynamically weighs branch outputs, and (iii) a DLinear complementary shortcut that provides full-window trend and seasonality context. MSMixer contains only 112K parameters at H=96 and runs at O(T) complexity. Evaluated on four ETT benchmarks with standard chronological splits and three random seeds, MSMixer achieves the lowest average MSE (0.357) among lightweight models, outperforming DLinear (0.386, -7.4%) and NLinear (0.365, -2.1%), winning 12 of 16 configurations. Against five Transformer-based baselines from the literature, MSMixer achieves best or second-best MSE in 9 of 16 configurations while using 5x fewer parameters than PatchTST. Ablation and sensitivity analyses confirm the complementary contributions of the multi-scale branches and the DLinear shortcut.