20260514:增加新内容
This commit is contained in:
33
concepts/fourier-filter-dynamics.md
Normal file
33
concepts/fourier-filter-dynamics.md
Normal file
@@ -0,0 +1,33 @@
|
||||
---
|
||||
title: "Fourier Filter for Dynamics(Fourier Filter 动力学分解)"
|
||||
created: 2026-05-11
|
||||
updated: 2026-05-11
|
||||
type: concept
|
||||
tags: [signal-processing, time-series, dynamics-decomposition]
|
||||
sources: [[liu-koopa-2023]]
|
||||
---
|
||||
|
||||
# Fourier Filter for Dynamics(Fourier Filter 动力学分解)
|
||||
|
||||
## 定义
|
||||
|
||||
Fourier Filter 是 Koopa 模型中用于解耦[[non-stationary-time-series|非平稳时间序列]]中时变与时不变分量的模块。通过在频域进行选择性滤波,将序列分解为两个动力学特性不同的子信号。
|
||||
|
||||
## 分解策略
|
||||
|
||||
| 分量 | 频域特性 | 动力学特性 | 处理方式 |
|
||||
|------|----------|------------|----------|
|
||||
| **时不变** | 低频 | 全局稳定、可长期预测 | 全局 Koopman 算子 |
|
||||
| **时变** | 高频 | 局部变化、强非平稳 | 上下文感知 Koopman 算子 |
|
||||
|
||||
## 工程意义
|
||||
|
||||
- 显式分离使[[koopman-predictor|Koopman 预测器]]可以**分别建模**两种动力学
|
||||
- 低频分量对应趋势/季节性,高频分量对应局部波动/突发事件
|
||||
- 频域操作用 FFT 实现,计算极高效
|
||||
|
||||
## 相关概念
|
||||
|
||||
- [[non-stationary-time-series|非平稳时间序列]]
|
||||
- [[time-variant-dynamics|时变动力学]]
|
||||
- [[koopman-predictor|Koopman 预测器]]
|
||||
Reference in New Issue
Block a user