跳转至

多值分析

For a normed space \((X, \|\cdot\|)\) ,let

\[ \begin{gather*} \mathscr{P}_{cl}(X) = \{ Y \in \mathscr{P}: Y \textsf{ is closed } \} \\ \mathscr{P}_{b}(X) = \{ Y \in \mathscr{P}: Y \textsf{ is bounded } \} \\ \mathscr{P}_{cp}(X) = \{ Y \in \mathscr{P}: Y \textsf{ is compact } \} \\ \mathscr{P}_{cp,c}(X) = \{ Y \in \mathscr{P}: Y \textsf{ is compact and convex } \} \\ \mathscr{P}_{cl, b}(X) = \{ Y \in \mathscr{P}: Y \textsf{ is closed and bounded } \} \end{gather*} \]

A multivalued map \(G: X \to \mathscr{P}(X)\)

3 . 称为 上半连续(upper semi-continuous (u.s.c.)), 若

@doc """
    usc(G::MultvalueMap, X::NormedSpace)

The upper semi-continuous function.

# Arguments
- `G::MultvalueMap`: A multivalued map.
- `X::NormedSpace`: A normed space.

"""
function usc(G::MultvalueMap, X::NormedSpace)
    for x0  X
        G(x0) is closed
        &&
        for N::OpenSet  X where G(x0)  N
             N0 where 
            x0  N0 
            &&
            G(N0)  N
        end
    end
end

4 . 称为 下半连续(lower semi-continuous (l.s.c.)), 若

@doc """
    lsc(G::MultvalueMap, X::NormedSpace)

The lower semi-continuous function.

# Arguments
- `G::MultvalueMap`: A multivalued map.
- `X::NormedSpace`: A normed space.

"""
function lsc(G::MultvalueMap, X::NormedSpace)
    for B::OpenSet  X
        Y := {y  X where G(y)  B  }
        Y is open
    end
end

Think

我们在凸优化中定义过所谓 下半连续。他是由单值函数的连续性来定义的。
而在多值函数中,没有了实数的 序关系连续性,那么应该用何种方式来替代呢?
在多值函数中,我们使用了 集合的序关系关系 以及由 开集 所描绘的连续性。
如果不理解,对于一个下半连续函数 \(f(x)\) ,不妨令 \( G(x) = \{y | y \le f(x)\}\),是不是就一目了然了。