SeesawNet: Towards Non-stationary Time Series Forecasting with Balanced Modeling of Common and Specific Dependencies

2026-05-14

AI Summary

This work addresses the challenge in non-stationary multivariate time series forecasting, where existing methods struggle to adaptively balance the modeling of shared (common) and instance-specific dependencies, and instance normalization often oversmooths unique structural patterns. To overcome these limitations, the authors propose SeesawNet, a novel architecture featuring an Adaptive Stationary-Nonstationary Attention (ASNA) mechanism that dynamically adjusts the fusion weights between normalized sequences (capturing commonality) and raw sequences (preserving specificity) based on instance-level non-stationarity. Additionally, SeesawNet incorporates alternating time-channel relational modeling modules to achieve dynamic dependency balancing across both temporal and channel dimensions. Extensive experiments demonstrate that the proposed method significantly outperforms state-of-the-art models on multiple real-world benchmarks, yielding substantial improvements in prediction accuracy.

Abstract

Instance normalization (IN) is widely used in non-stationary multivariate time series forecasting to reduce distribution shifts and highlight common patterns across samples. However, IN can over-smooth instance-specific structural information that is essential for modeling temporal and cross-channel heterogeneity. While prior methods further suppress distribution discrepancies or attempt to recover temporal specific dependencies, they often ignore a central tension: how to adaptively model common and instance-specific dependency based on each instance's non-stationary structures. To address this dilemma, we propose SeesawNet, a unified architecture that dynamically balances common and instance-specific dependency modeling in both temporal and channel dimensions. At its core is Adaptive Stationary-Nonstationary Attention (ASNA), which captures common dependencies from normalized sequences and specific dependencies from raw sequences, and adaptively fuses them according to instance-level non-stationarity. Built upon ASNA, SeesawNet alternates dedicated temporal and channel relationship modeling to jointly capture long-range and cross-variable dependencies. Extensive experiments on multiple real-world benchmarks demonstrate that SeesawNet consistently outperforms state-of-the-art methods.

Problem

• non-stationary time series forecasting
• common dependencies
• instance-specific dependencies
• temporal heterogeneity
• channel heterogeneity

Innovation

• SeesawNet
• non-stationary time series forecasting
• instance-specific dependency
• adaptive attention
• multivariate time series