针对周期性数据建模弱的问题，提出频谱感知的神经过程模型。
Spectral Transformer Neural Processes
- 引入频谱聚合器，将上下文数据转为频谱混合特征
- 在合成与真实时间序列上预测性能显著优于基线
- 适合处理具有周期或准周期特性的时序与图像数据
时间序列、空间数据和图像是神经过程的自然应用场景。然而，当这些数据表现出强周期性或准周期性时，现有方法常出现欠拟合且泛化能力差。本文提出频谱感知的Transformer神经过程（STNPs），在Transformer神经过程基础上引入频谱聚合器，通过估计上下文数据的实证频谱，将其压缩为频谱混合，采样任务自适应的频谱特征，并与时间域嵌入拼接，从而向神经过程注入频谱混合核偏置。该设计重塑了相似性几何结构，使欧氏距离较远的输入在诱导的周期性流形中仍保持相近，同时增强时频交互。在合成回归任务、真实世界时间序列数据集及图像数据集上的大量实验表明，STNPs在预测性能上持续优于现有基线，使神经过程从平移等变性扩展至有效建模周期性和准周期性数据。
原文摘要 · Abstract (English)
Time series, spatial data, and images are natural applications of Neural Processes. However, when such data exhibit strong periodicity and quasi-periodicity, existing methods often suffer from underfitting and generalise poorly beyond the training distribution. In this work, we propose Spectral Transformer Neural Processes (STNPs), a frequency-aware extension of Transformer Neural Processes (TNPs). STNPs introduce a Spectral Aggregator that estimates an empirical context spectrum, compresses it into a spectral mixture, samples task-adaptive spectral features, and concatenates them with time-domain embeddings, thereby injecting a spectral-mixture-kernel bias into TNPs. This design reshapes the similarity geometry, allowing inputs that are distant in Euclidean space to remain close in an induced periodic manifold while enhancing time-frequency interactions. Extensive experiments on synthetic regression tasks, real-world time-series datasets, and an image dataset demonstrate that STNPs consistently improve predictive performance over existing baselines, extending Neural Processes beyond translation equivariance towards effective modelling of periodicity and quasi-periodicity.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。