Sequential Order-Robust Mamba for Time Series Forecasting
This work addresses a problem in time series forecasting for researchers and practitioners by mitigating channel order bias in Mamba models, though it is incremental as it builds on existing Mamba frameworks.
The paper tackles the issue of sequential order bias in Mamba-based time series forecasting by proposing SOR-Mamba, which incorporates regularization for channel order robustness and eliminates 1D-convolution, resulting in improved performance across standard and transfer learning scenarios.
Mamba has recently emerged as a promising alternative to Transformers, offering near-linear complexity in processing sequential data. However, while channels in time series (TS) data have no specific order in general, recent studies have adopted Mamba to capture channel dependencies (CD) in TS, introducing a sequential order bias. To address this issue, we propose SOR-Mamba, a TS forecasting method that 1) incorporates a regularization strategy to minimize the discrepancy between two embedding vectors generated from data with reversed channel orders, thereby enhancing robustness to channel order, and 2) eliminates the 1D-convolution originally designed to capture local information in sequential data. Furthermore, we introduce channel correlation modeling (CCM), a pretraining task aimed at preserving correlations between channels from the data space to the latent space in order to enhance the ability to capture CD. Extensive experiments demonstrate the efficacy of the proposed method across standard and transfer learning scenarios. Code is available at https://github.com/seunghan96/SOR-Mamba.