Skip to content

Commit

Permalink
feat: Add functional analysis notes
Browse files Browse the repository at this point in the history
  • Loading branch information
tiankaima committed Jul 27, 2024
1 parent 90435f1 commit 5059eff
Show file tree
Hide file tree
Showing 2 changed files with 126 additions and 0 deletions.
125 changes: 125 additions & 0 deletions docs/math/analysis/functional-analysis.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,125 @@
# 泛函分析

## 常见空间

### $l_p$ 空间 {#countable-lp-norm}

$l_p$ 空间的定义:

$$
l_p = \left\{\mathbf{x} = (x_1, x_2, \ldots) \left| \sum_{i=1}^{\infty} \left| x_i \right|^p < \infty\right.\right\}
$$

### $L^p$ 空间 {#lp-space}

$L^p$ 空间的定义:

$$
\mathcal{L}ƒ^p(X,\mu) = \left\{f: X \to \mathbb{R} \left| \int_X \left| f(x) \right|^p \mathrm{d}x < \infty\right.\right\}
$$

???+ note "等价关系"

严格来说,上面的定义并不是 $L^p(X,\mu)$, 而是 $\mathcal{L}^p(X,\mu)$, 因为 $\lVert \cdot \rVert_p$ 只构成一个半范数,而不是范数。采取一个较为标准的拓扑方法,只需要考虑 $\mathcal{L}^p(X,\mu)$ 满足 $\lVert f \rVert_p = 0$ 的全部函数:

$$
N = \left\{f: X \to \mathbb{R} \left| \int_X \left| f(x) \right|^p \mathrm{d}x = 0\right.\right\}
$$

即 $f \rightarrow \lVert f \rVert_p$ 的零空间。对于可测函数来说 $\lVert f\rVert_p = 0 \Leftrightarrow \mu \left(f\neq 0\right) = 0$. 只需要将 $L_p(X,\mu)$ 定义为 $\mathcal{L}^p(X,\mu)/N$ 即可。

## Fubini / Tonelli 定理

- Fubini 定理:考虑 $f(x,y)$ 在 $X\times Y$ 的一个矩形区域上 Lebesgue 可积,即

$$
\int_{X\times Y} \left| f(x,y) \right| \mathrm{d}(x,y) < \infty
$$

那么:

$$
\int_{X\times Y} f(x,y) \mathrm{d}(x,y) = \int_X \left(\int_Y f(x,y) \mathrm{d}y\right) \mathrm{d}x = \int_Y \left(\int_X f(x,y) \mathrm{d}x\right) \mathrm{d}y
$$

- Tonelli 定理:考虑 $f(x,y)$ 在 $X\times Y$ 的一个矩形区域上非负可测,即:$f: X\times Y \to [0, \infty]$, 那么:

$$
\int_{X\times Y} f(x,y) \mathrm{d}x\mathrm{d}y = \int_X \left(\int_Y f(x,y) \mathrm{d}y\right) \mathrm{d}x = \int_Y \left(\int_X f(x,y) \mathrm{d}x\right) \mathrm{d}y
$$

## 卷积 Convolution

### 定义

- 连续情况:

考虑 $L^1$ 上的两个函数 $f(x)$ 和 $g(x)$,它们的卷积定义为:

$$
(f * g)(x) = \int_{-\infty}^{\infty} f(x - t) g(t) \mathrm{d} t
$$

!!! note ""

考虑定义域限制 $f, g: [0,\infty) \to \mathbb{R}$,则卷积操作变为:

$$
(f * g)(x) = \int_{0}^{x} f(x - t) g(t) \mathrm{d}t
$$

> 实际上是上面定义的自然推广,只需补充 $f, g: (-\infty,0) \to \{0\}$ 即可。

一般的,对于 $L^1\left(\mathbb{R}^n\right)$ 上的函数,卷积定义为:

$$
(f * g)(\mathbf{x}) = \int_{\mathbb{R}^n} f(\mathbf{x} - \mathbf{t}) g(\mathbf{t}) \mathrm{d}\mathbf{t}
$$

- 离散情况:

考虑 $l^1$ 上的两个序列 $\{f_n\}$ 和 $\{g_n\}$,它们的卷积定义为:

$$
(f * g)_n = \sum_{k=-\infty}^{\infty} f_{n-k} \cdot g_k
$$

### 性质

- 交换性,即 $(f * g)(x) = (g * f)(x)$

- 对 $f, g\in L^1\left(\mathbb{R}^n\right)$, $f*g$ 存在,而且 $f*g \in L^1\left(\mathbb{R}^n\right)$

- 更一般的, $f\in L^1\left(\mathbb{R}^n\right)$, $g\in L^p\left(\mathbb{R}^n\right)$, $1\leq p\leq \infty$, 则 $f*g \in L^p\left(\mathbb{R}^n\right)$, 且有:

$$
\left\| f*g \right\|_p \leq \left\| f \right\|_1 \left\| g \right\|_p
$$

- 与 Fourier 变换 之间的联系:

$$
\mathcal{F}\left[f*g\right] = \mathcal{F}\left[f\right] \cdot \mathcal{F}\left[g\right]
$$

- 微分/积分特性,下面 $h = f * g$

$$
h'(t) = \left(f * g\right)'(t) = f'(t) * g(t) = f(t) * g'(t)
$$

$$
\int_{-\infty}^{\infty} h(t) \mathrm{d}t = \left(\int_{-\infty}^{\infty} f(t) \mathrm{d}t\right) * g(t) = f(t) * \left(\int_{-\infty}^{\infty} g(t) \mathrm{d}t\right)
$$

$$
h(t) = f'(t) * \left(\int_{-\infty}^{\infty} g(t) \mathrm{d}t\right) = \left(\int_{-\infty}^{\infty} f(t) \mathrm{d}t\right) * g'(t)
$$

- 特别的,考虑 $\delta (t)$:

- $f(t) * \delta(t) = f(t)$
- $f(t) * \delta(t - t_0) = f(t - t_0)$
- $f(t) * \delta'(t) = f'(t) * \delta(t) = f'(t)$
- $f(t) * \delta^{(n)}(t) = f^{(n)}(t) * \delta(t) = f^{(n)}(t)$
- $\displaystyle f(t) * \int_{-\infty}^{t} \delta(t) \mathrm{d}t = \int_{-\infty}^{t} f(t) \mathrm{d}t = F(t)$
1 change: 1 addition & 0 deletions mkdocs.yml
Original file line number Diff line number Diff line change
Expand Up @@ -31,6 +31,7 @@ nav:
- math/index.md
- Analysis:
- math/analysis/calculus.md
- math/analysis/functional-analysis.md
- Probability:
- math/probability/probability.md
- math/probability/stochastic-processes.md
Expand Down

0 comments on commit 5059eff

Please sign in to comment.