Kernel-U-Net: Multivariate Time Series Forecasting using Custom Kernels
This work addresses computational efficiency and expressiveness issues in time series forecasting for researchers and practitioners, though it appears incremental as it builds on existing U-Net and kernel methods.
The paper tackles the limitations of Transformer-based U-Net architectures in multivariate time series forecasting by introducing Kernel-U-Net, a flexible and kernel-customizable U-shape neural network, which achieves performance that exceeds or meets state-of-the-art models on seven real-world datasets in channel-independent settings.
Time series forecasting task predicts future trends based on historical information. Transformer-based U-Net architectures, despite their success in medical image segmentation, have limitations in both expressiveness and computation efficiency in time series forecasting as evidenced in YFormer. To tackle these challenges, we introduce Kernel-U-Net, a flexible and kernel-customizable U-shape neural network architecture. The kernel-U-Net encoder compresses the input series into latent vectors, and its symmetric decoder subsequently expands these vectors into output series. Specifically, Kernel-U-Net separates the procedure of partitioning input time series into patches from kernel manipulation, thereby providing the convenience of customized executing kernels. Our method offers two primary advantages: 1) Flexibility in kernel customization to adapt to specific datasets; and 2) Enhanced computational efficiency, with the complexity of the Transformer layer reduced to linear. Experiments on seven real-world datasets, demonstrate that Kernel-U-Net's performance either exceeds or meets that of the existing state-of-the-art model in the majority of cases in channel-independent settings. The source code for Kernel-U-Net will be made publicly available for further research and application.