DualWeaver: 用于基于单变量时间序列基础模型的多变量预测的协同特征编织替代方案
DualWeaver: Synergistic Feature Weaving Surrogates for Multivariate Forecasting with Univariate Time Series Foundation Models
AI总结：
DualWeaver通过结构对称的替代序列提升多变量预测的准确性和稳定性，利用共享特征融合模块和无参数解码实现高效预测。
摘要：
时间序列基础模型（TSFMs）通过大规模预训练实现了强大的单变量预测，但有效将这一成功扩展到多变量预测仍具挑战性。为此，我们提出了DualWeaver，一种新颖的框架，通过使用一对可学习的、结构上对称的替代序列，将单变量TSFMs（Uni-TSFMs）适应于多变量预测。这些替代序列由一个共享的辅助特征融合模块生成，该模块捕捉跨变量依赖关系，然后通过预测目标映射到TSFM兼容的序列。对称结构使能够无参数地从替代序列直接重建最终预测，而无需额外参数解码。进一步引入了一个理论支撑的正则化项，以增强对适应崩溃的鲁棒性。在多样化的现实世界数据集上的广泛实验表明，DualWeaver在准确性和稳定性方面均优于最先进的多变量预测器。我们将在https://github.com/li-jinpeng/DualWeaver上发布代码。
英文摘要：
Time-series foundation models (TSFMs) have achieved strong univariate forecasting through large-scale pre-training, yet effectively extending this success to multivariate forecasting remains challenging. To address this, we propose DualWeaver, a novel framework that adapts univariate TSFMs (Uni-TSFMs) for multivariate forecasting by using a pair of learnable, structurally symmetric surrogate series. Generated by a shared auxiliary feature-fusion module that captures cross-variable dependencies, these surrogates are mapped to TSFM-compatible series via the forecasting objective. The symmetric structure enables parameter-free reconstruction of final predictions directly from the surrogates, without additional parametric decoding. A theoretically grounded regularization term is further introduced to enhance robustness against adaptation collapse. Extensive experiments on diverse real-world datasets show that DualWeaver outperforms state-of-the-art multivariate forecasters in both accuracy and stability. We release the code at https://github.com/li-jinpeng/DualWeaver.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
