측도론 에서 가측 함수 (可測函數, 영어 : measurable function )는 원상 이 가측성을 보존하는 함수이다.
두 가측 공간
(
X
,
F
)
{\displaystyle (X,{\mathcal {F}})}
,
(
Y
,
G
)
{\displaystyle (Y,{\mathcal {G}})}
사이의 가측 함수
f
:
X
→
Y
{\displaystyle f\colon X\to Y}
는 다음 성질을 만족시키는 함수이다.
모든
S
∈
G
{\displaystyle S\in {\mathcal {G}}}
에 대하여,
f
−
1
(
S
)
∈
F
{\displaystyle f^{-1}(S)\in {\mathcal {F}}}
만약 공역 이 유클리드 공간 인 경우, 보통 공역에 보렐 시그마 대수 를 부여한다. 만약 정의역 이 유클리드 공간 일 영우, 보통 공역에 르베그 시그마 대수 를 부여한다. 즉, "가측 함수
R
→
R
{\displaystyle \mathbb {R} \to \mathbb {R} }
"는 보통
(
R
,
L
(
R
)
)
→
(
R
,
B
(
R
)
)
{\displaystyle (\mathbb {R} ,{\mathcal {L}}(\mathbb {R} ))\to (\mathbb {R} ,{\mathcal {B}}(\mathbb {R} ))}
을 의미한다.
두 가측 함수
f
:
(
X
1
,
F
1
)
→
(
X
2
,
F
2
)
{\displaystyle f\colon (X_{1},{\mathcal {F}}_{1})\to (X_{2},{\mathcal {F}}_{2})}
g
:
(
X
2
,
F
2
)
→
(
X
3
,
F
3
)
{\displaystyle g\colon (X_{2},{\mathcal {F}}_{2})\to (X_{3},{\mathcal {F}}_{3})}
가 주어졌을 때, 그 합성 함수
g
∘
f
:
(
X
1
,
F
1
)
→
(
X
3
,
F
3
)
{\displaystyle g\circ f\colon (X_{1},{\mathcal {F}}_{1})\to (X_{3},{\mathcal {F}}_{3})}
역시 가측 함수이다.
X
{\displaystyle X}
와
Y
{\displaystyle Y}
가 보렐 시그마 대수 를 갖춘 위상 공간 이라고 하면, 다음이 성립한다.
X
{\displaystyle X}
와
Y
{\displaystyle Y}
사이의 모든 연속 함수 는 가측 함수이다.
반대로, 루진의 정리 에 따르면,
Y
{\displaystyle Y}
가 제2 가산 공간 이며
X
{\displaystyle X}
에 라돈 측도 가 부여되었다면, 모든 가측 함수
f
:
X
→
Y
{\displaystyle f\colon X\to Y}
는
X
{\displaystyle X}
의 (라돈 측도에 대하여) 거의 어디서나 연속 함수이다.
(
X
,
F
)
{\displaystyle (X,{\mathcal {F}})}
가 임의의 가측 공간일 경우, 다음이 성립한다.
두 가측 함수
f
,
g
:
(
X
,
F
)
→
(
R
,
B
(
R
)
)
{\displaystyle f,g\colon (X,{\mathcal {F}})\to (\mathbb {R} ,{\mathcal {B}}(\mathbb {R} ))}
에 대하여,
f
+
g
{\displaystyle f+g}
및
f
⋅
g
{\displaystyle f\cdot g}
는 가측 함수이다.
가측 함수의 열
f
1
,
f
2
,
…
:
(
X
,
F
)
→
(
R
,
B
(
R
)
)
{\displaystyle f_{1},f_{2},\dots \colon (X,{\mathcal {F}})\to (\mathbb {R} ,{\mathcal {B}}(\mathbb {R} ))}
의 점별 극한은 가측 함수이다.
모든 르베그 적분 가능 함수
X
→
R
{\displaystyle X\to \mathbb {R} }
는 가측 함수이다.
임의의 함수
f
:
R
→
R
{\displaystyle f\colon \mathbb {R} \to \mathbb {R} }
및
g
:
R
→
[
0
,
∞
)
{\displaystyle g\colon \mathbb {R} \to [0,\infty )}
에 대하여, 다음 두 조건이 서로 동치 이다.
f
:
(
R
,
L
(
R
)
)
→
(
R
,
B
(
R
)
)
{\displaystyle f\colon (\mathbb {R} ,{\mathcal {L}}(\mathbb {R} ))\to (\mathbb {R} ,{\mathcal {B}}(\mathbb {R} ))}
는 가측 함수이다.
다음 함수는 르베그 적분 가능 함수이다.
mid
{
−
g
,
f
,
g
}
:
x
↦
{
g
(
x
)
f
(
x
)
≥
g
(
x
)
f
(
x
)
−
g
(
x
)
≤
f
(
x
)
≤
g
(
x
)
−
g
(
x
)
f
(
x
)
≤
−
g
(
x
)
{\displaystyle \operatorname {mid} \{-g,f,g\}\colon x\mapsto {\begin{cases}g(x)&f(x)\geq g(x)\\f(x)&-g(x)\leq f(x)\leq g(x)\\-g(x)&f(x)\leq -g(x)\end{cases}}}
다음 데이터가 주어졌다고 하자.
가측 공간
(
X
,
F
)
{\displaystyle (X,{\mathcal {F}})}
K
∈
{
R
,
C
}
{\displaystyle \mathbb {K} \in \{\mathbb {R} ,\mathbb {C} \}}
(표준적인 위상과 보렐 시그마 대수 를 갖춘)
K
{\displaystyle \mathbb {K} }
-바나흐 공간
Y
{\displaystyle Y}
그렇다면,
X
→
Y
{\displaystyle X\to Y}
단순 함수 는 다음과 같은 꼴의 함수
f
:
X
→
Y
{\displaystyle f\colon X\to Y}
이다.
f
=
∑
i
=
1
k
y
i
1
S
i
{\displaystyle f=\sum _{i=1}^{k}y_{i}1_{S_{i}}}
y
1
,
…
,
y
k
∈
Y
{\displaystyle y_{1},\dots ,y_{k}\in Y}
S
1
,
…
,
S
k
∈
F
{\displaystyle S_{1},\dots ,S_{k}\in {\mathcal {F}}}
k
∈
N
{\displaystyle k\in \mathbb {N} }
(여기서
1
S
i
{\displaystyle 1_{S_{i}}}
는 지시 함수 이다.)
이제, 함수
f
:
X
→
Y
{\displaystyle f\colon X\to Y}
에 대하여, 다음 세 조건을 정의하자.
(강가측 함수 , 强可測函數, 영어 : strongly measurable function ) 단순 함수 의 열의 점별 극한이다.
(약가측 함수 , 弱可測函數, 영어 : weakly measurable function ) 임의의 연속 쌍대 공간 원소
ϕ
∈
Y
∗
{\displaystyle \phi \in Y^{*}}
에 대하여,
ϕ
∘
f
:
(
X
,
B
(
X
)
)
→
(
K
,
B
(
K
)
)
{\displaystyle \phi \circ f\colon (X,{\mathcal {B}}(X))\to (\mathbb {K} ,{\mathcal {B}}(\mathbb {K} ))}
는 가측 함수이다.
이 경우, 모든 강가측 함수는 가측 함수이며, 모든 가측 함수는 약가측 함수이다. 또한, 페티스 가측성 정리 (Pettis可測性定理, 영어 : Pettis measurability theorem )에 따르면, 함수
f
:
X
→
Y
{\displaystyle f\colon X\to Y}
에 대하여, 다음 두 조건이 서로 동치 이다.[ 1] :5, Theorem 1.1.6
강가측 함수이다.
약가측 함수이며,
f
(
X
)
⊂
Y
~
{\displaystyle f(X)\subset {\widetilde {Y}}}
인 분해 가능 부분 공간
Y
~
⊂
Y
{\displaystyle {\widetilde {Y}}\subset Y}
가 존재한다.
특히, 만약
Y
{\displaystyle Y}
가 분해 가능 바나흐 공간 일 경우,
f
:
X
→
Y
{\displaystyle f\colon X\to Y}
에 대하여, 다음 세 조건이 서로 동치 이다.
강가측 함수이다.
가측 함수이다.
약가측 함수이다.
가측 공간
(
X
,
F
)
{\displaystyle (X,{\mathcal {F}})}
및
K
∈
{
R
,
C
}
{\displaystyle \mathbb {K} \in \{\mathbb {R} ,\mathbb {C} \}}
및 두
K
{\displaystyle \mathbb {K} }
-바나흐 공간
Y
{\displaystyle Y}
,
Z
{\displaystyle Z}
가 주어졌다고 하자. 만약
f
:
X
→
Y
{\displaystyle f\colon X\to Y}
가 강가측 함수이며,
g
:
Y
→
Z
{\displaystyle g\colon Y\to Z}
가 가측 함수라면,
g
∘
f
{\displaystyle g\circ f}
는 강가측 함수이다.[ 1] :7, Corollary 1.1.11
정의에 따르면 확률 변수 는 확률 공간 을 정의역으로 하는 가측 함수이다.
모든 함수가 가측 함수는 아니다. 예를 들면, 만약
A
⊂
R
{\displaystyle A\subset \mathbb {R} }
가 가측 집합이 아닌 경우, 지시 함수
1
A
{\displaystyle 1_{A}}
는 가측 함수가 아니다.
분해 불가능
K
{\displaystyle \mathbb {K} }
-바나흐 공간
X
{\displaystyle X}
위의 항등 함수
f
:
x
↦
x
{\displaystyle f\colon x\mapsto x}
를 생각하자. 이는 연속 함수 이므로 보렐 가측 함수이다. 그러나
f
(
X
)
=
X
{\displaystyle f(X)=X}
가 분해 가능하지 않으므로, 페티스 가측성 정리에 따라
f
{\displaystyle f}
는 강가측 함수가 아니다.[ 1] :4, Example 1.1.5
실수의 셈측도 공간
(
R
,
P
(
R
)
,
μ
)
{\displaystyle (\mathbb {R} ,{\mathcal {P}}(\mathbb {R} ),\mu )}
위의 르베그 공간
ℓ
2
(
R
;
K
)
{\displaystyle \ell ^{2}(\mathbb {R} ;\mathbb {K} )}
를 공역으로 하는, 다음과 같은 함수를 정의하자.
f
:
R
→
ℓ
2
(
R
;
K
)
{\displaystyle f\colon \mathbb {R} \to \ell ^{2}(\mathbb {R} ;\mathbb {K} )}
f
(
x
)
(
t
)
=
{
1
t
=
x
0
t
≠
x
{\displaystyle f(x)(t)={\begin{cases}1&t=x\\0&t\neq x\end{cases}}}
그렇다면,
f
:
(
R
,
L
(
R
)
)
→
(
ℓ
2
(
R
;
K
)
,
B
(
ℓ
2
(
R
;
K
)
)
)
{\displaystyle f\colon (\mathbb {R} ,{\mathcal {L}}(\mathbb {R} ))\to (\ell ^{2}(\mathbb {R} ;\mathbb {K} ),{\mathcal {B}}(\ell ^{2}(\mathbb {R} ;\mathbb {K} )))}
는 약가측 함수이지만, 가측 함수가 아니다. 우선, 임의의
x
∈
R
{\displaystyle x\in \mathbb {R} }
에 대하여,
⟨
f
,
f
(
x
)
⟩
=
f
(
x
)
:
(
R
,
L
(
R
)
)
→
(
K
,
B
(
K
)
)
{\displaystyle \langle f,f(x)\rangle =f(x)\colon (\mathbb {R} ,{\mathcal {L}}(\mathbb {R} ))\to (\mathbb {K} ,{\mathcal {B}}(\mathbb {K} ))}
는 가측 함수이다. 또한, 비가측 집합
A
⊂
R
{\displaystyle A\subset \mathbb {R} }
에 대하여,
B
=
⋃
x
∈
A
ball
ℓ
2
(
R
;
K
)
(
1
,
f
(
x
)
)
⊂
ℓ
2
(
R
;
K
)
{\displaystyle B=\bigcup _{x\in A}\operatorname {ball} _{\ell ^{2}(\mathbb {R} ;\mathbb {K} )}(1,f(x))\subset \ell ^{2}(\mathbb {R} ;\mathbb {K} )}
는 열린집합 이므로 가측 집합이지만, 그 원상
f
−
1
(
B
)
=
A
{\displaystyle f^{-1}(B)=A}
는 가측 집합이 아니다.