📄 中文摘要
标准的直接预测模型通常依赖于均方误差等点对点目标，无法捕捉图结构信号中固有的复杂时空依赖性。虽然最近的频域方法如FreDF缓解了时间自相关问题，但往往忽视了空间和跨时空交互的影响。为了解决这一局限，提出了FreST Loss，这是一种增强频域的时空训练目标，扩展了对联合时空谱的监督。通过利用联合傅里叶变换（JFT），FreST Loss在统一的谱域中对齐模型预测与真实值，有效地去相关复杂的时空依赖性。
去相关化未来：联合频域学习用于时空预测
📄 English Summary
Decorrelating the Future: Joint Frequency Domain Learning for Spatio-temporal Forecasting
Standard direct forecasting models typically rely on point-wise objectives such as Mean Squared Error, which fail to capture the complex spatio-temporal dependencies inherent in graph-structured signals. Recent frequency-domain approaches like FreDF mitigate temporal autocorrelation but often overlook spatial and cross spatio-temporal interactions. To address this limitation, FreST Loss is proposed as a frequency-enhanced spatio-temporal training objective that extends supervision to the joint spatio-temporal spectrum. By leveraging the Joint Fourier Transform (JFT), FreST Loss aligns model predictions with ground truth in a unified spectral domain, effectively decorrelating complex dependencies across both space and time.