Machine learningTime-series forecasting
FEDformer:频率增强分解Transformer
FEDformer是由Zhou等人于2022年在ICML上提出的,一种基于Transformer的长期多元时间序列预测架构。其核心创新在于将季节-趋势分解与频域注意力相结合:FEDformer不计算时域中的完整token-to-token注意力,而是通过傅里叶变换或小波变换将查询(queries)、键(keys)和值(values)投影到频域,并在随机选择的频率分量子集上进行操作,从而实现线性复杂度,同时保留全局时间结构。
阅读完整方法
仅限会员
登录使用免费账户登录即可阅读本节。
Method map
The neighbourhood of related methods — select a node to explore.
来源
- Zhou, T., Ma, Z., Wen, Q., Wang, X., Sun, L., & Jin, R. (2022). FEDformer: Frequency enhanced decomposed transformer for long-term series forecasting. ICML. link ↗
如何引用本页
ScholarGate. (2026, June 2). FEDformer (Frequency Enhanced Decomposed Transformer). ScholarGate. https://scholargate.app/zh/deep-learning/fedformer
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Autoformer:用于长期时间序列预测的分解Transformer深度学习↔ compare
- FiLM: 频率改进的勒让德记忆模型深度学习↔ compare
- Informer深度学习↔ compare