Appendix A Overview on Suprema and Limits
Appendix A Overview on Suprema and Limits
A. 1 Supremum and maximum
这里我们回顾了有关上确界和极限的基本结果,这些结果对于发展信息理论编码定理非常有用。
- 我们的讨论始终在 \(\mathbb{R}\) 的子集上进行,即实数集。
定义 A.1(集合的上界)如果一个实数 \(u\) 是非空子集 \(\mathcal{A}\) 的上界,即 \(\mathcal{A}\) 中的每一个元素都小于或等于 \(u\),我们说 \(\mathcal{A}\) 是有上界的。符号上,定义表示为:
\(\mathcal{A} {\subset} \mathbb{R}\) 有上界 \({\Leftrightarrow} \left( {\exists}u {\in} \mathbb{R} \right)\) 使得 \(\left( {\forall}a {\in} \mathcal{A} \right),a {\leq} u\)。
Definition A. 1 (Upper bound of a set) A real number \(u\) is called an upper bound of a non-empty subset \(\mathcal{A}\) of \(\mathbb{R}\) if every element of \(\mathcal{A}\) is less than or equal to \(u\) ; we say that \(\mathcal{A}\) is bounded above. Symbolically,the definition becomes:
\(\mathcal{A} {\subset} \mathbb{R}\) is bounded above \({\Leftrightarrow} \left( {\exists}u {\in} \mathbb{R} \right)\) such that \(\left( {\forall}a {\in} \mathcal{A} \right),a {\leq} u\) .
定义 A.2(最小上界或上确界)假设 \(\mathcal{A}\) 是 \(\mathbb{R}\) 的一个非空子集。如果实数 \(s\) 是 \(\mathcal{A}\) 的上界,并且对于 \(\mathcal{A}\) 的任何其他上界 \(s^{{\prime}}\),都有 \(s {\leq} s^{{\prime}}\),那么我们说实数 \(s\) 是 \(\mathcal{A}\) 的最小上界或上确界。在这种情况下,我们写 \(s = \sup\mathcal{A}\);其他表示方法包括 \(s = {\sup}_{x {\in} \mathcal{A}}x\) 和 \(s = \sup\{x : x {\in} \mathcal{A}\}\)。
Definition A. 2 (Least upper bound or supremum) Suppose \(\mathcal{A}\) is a non-empty subset of \(\mathbb{R}\) . Then we say that a real number \(s\) is a least upper bound or supremum of \(\mathcal{A}\) if \(s\) is an upper bound of the set \(\mathcal{A}\) and if \(s {\leq} s^{{\prime}}\) for each upper bound \(s^{{\prime}}\) of \(\mathcal{A}\) . In this case,we write \(s = \sup\mathcal{A}\) ; other notations are \(s = {\sup}_{x {\in} \mathcal{A}}x\) and \(s = \sup\{x : x {\in} \mathcal{A}\}\) .
完备性公理:(最小上界性质)让 \(\mathcal{A}\) 是一个有上界的非空子集 \(\mathbb{R}\)。那么 \(\mathcal{A}\) 存在最小上界(在 \(\mathbb{R}\) 中)。
Completeness Axiom: (Least upper bound property) Let \(\mathcal{A}\) be a non-empty subset of \(\mathbb{R}\) that is bounded above. Then \(\mathcal{A}\) has a least upper bound (in \(\mathbb{R})\)
完备性公理可能不适用于,比如,\(\mathbb{Q}\),有理数集。
定义 A.1'(有理数集中的上界)如果有理数 \(u\) 是非空子集 \(\mathcal{A}\) 的上界,即 \(\mathcal{A}\) 中的每一个元素都小于或等于 \(u\),我们说 \(\mathcal{A}\) 是有上界的。符号上,定义表示为:
\(\mathcal{A} {\subset} \mathbb{Q}\) 有上界 \({\Leftrightarrow} \left( {\exists}u {\in} \mathbb{Q} \right)\) 使得 \(\left( {\forall}a {\in} \mathcal{A} \right),a {\leq} u\)。
定义-A.2′(有理数集中的最小上界或上确界)假设 \(\mathcal{A}\) 是 \(\mathbb{Q}\) 的一个非空子集。如果有理数 \(s\) 是 \(\mathcal{A}\) 的上界,并且对于 \(\mathcal{A}\) 的任何其他上界 \(s^{{\prime}}\),都有 \(s {\leq} s^{{\prime}}\),那么我们说有理数 \(s\) 是 \(\mathcal{A}\) 的最小上界或上确界。在这种情况下,我们写 \(s = \sup\mathcal{A}\);其他表示方法包括 \(s = \mathop{\sup}\limits_{x {\in} \mathcal{A}}x\) 和 \(s = \sup\{x : x {\in} \mathcal{A}\}\)。
示例。\(\mathcal{A} = \left\{x {\in} \mathbb{Q} : x^{2} < 2 \right\}\)。那么,\(\mathop{\sup}\limits_{x {\in} \mathcal{A}}x\) 应该是“小于 \(\sqrt{2}\) 的最大有理数”!
A. 1 Supremum and maximum
-
It follows directly that if a non-empty set in \(\mathbb{R}\) has a supremum,then this supremum is unique.
-
By definition,the empty set \(({\varnothing})\) and any set not bounded above do not admit a supremum in \(\mathbb{R}\) .
-
直接推论,如果 \(\mathbb{R}\) 中的一个非空集合有上确界,那么这个上确界是唯一的。
-
根据定义,空集 \(({\varnothing})\) 和任何没有上界的集合在 \(\mathbb{R}\) 中都不存在上确界。
Property A.4 (Properties of the supremum)
- 在 \(\mathbb{R} {\cup} \{{-} {\infty},{\infty}\}\)(扩展的实数集)中任何集合的上确界总是存在的。
这些扩展定义将在本课程中采用。
- The supremum of any set in \(\mathbb{R} {\cup} \{{-} {\infty},{\infty}\}\) (the set of extended real numbers) always exits.
These extended definitions will be adopted in this course.
- \(\left( {\forall}a {\in} \mathcal{A} {\subset} \mathbb{R}{\cup}\{{-} {\infty},{\infty}\} \right)a {\leq} \sup\mathcal{A}\) 。
- 如果 \({-} {\infty} < \sup\mathcal{A} < {\infty}\) ,则对于任何 \(\varepsilon > 0\),都存在 \(a_{0} {\in} \mathcal{A}\) 使得 \(a_{0} > \sup\mathcal{A} {-} \varepsilon\) 。
(在 \(\left| \sup\mathcal{A} \right| < {\infty}\) 的条件下,存在 \(a_{0} {\in} (\sup\mathcal{A} {-} \varepsilon,\sup\mathcal{A}\rbrack\) 被称为上确界的逼近性质。)
- If \({-} {\infty} < \sup\mathcal{A} < {\infty}\) ,then \(({\forall}\varepsilon > 0)\left( {\exists}a_{0} {\in} \mathcal{A} \right)a_{0} > \sup\mathcal{A} {-} \varepsilon\) .
(The existence of \(a_{0} {\in} (\sup\mathcal{A} {-} \varepsilon,\sup\mathcal{A}\rbrack\) for any \(\varepsilon > 0\) under the condition of \(\left| \sup\mathcal{A} \right| < {\infty}\) is called the approximation property for the supremum.)
-
如果 \(\sup\mathcal{A} = {\infty}\) ,则对于任何 \(L {\in} \mathbb{R}\),都存在 \(B_{0} {\in} \mathcal{A}\) 满足 \(B_{0} > L\) 。
-
如果 \(\sup\mathcal{A} = {-} {\infty}\) ,则 \(\mathcal{A}\) 为空集。
定义 A. 3(最大值)如果 \(\sup\mathcal{A} {\in} \mathcal{A}\) ,则 \(\sup\mathcal{A}\) 也称为 \(\mathcal{A}\) 的最大值,并用 \(\max\mathcal{A}\) 表示。然而,如果 \(\sup\mathcal{A} {\notin} \mathcal{A}\),则我们说 \(\mathcal{A}\) 的最大值不存在。
例如,如果 \(\mathcal{A} = (0,1\rbrack\),则 \(\max\mathcal{A} = \sup A = 1\)。
例如,如果 \(\mathcal{A} = (0,1)\),则 \(\sup\mathcal{A} = 1\) 但 \(\max\mathcal{A}\) 不存在!
Definition A. 3 (Maximum) If \(\sup\mathcal{A} {\in} \mathcal{A}\) ,then \(\sup\mathcal{A}\) is also called the max-imum of \(\mathcal{A}\) ,and is denoted by \(\max\mathcal{A}\) . However,if \(\sup\mathcal{A} {\notin} \mathcal{A}\) ,then we say that the maximum of \(\mathcal{A}\) does not exist.
E.g.,if \(\mathcal{A} = (0,1\rbrack\) ,then \(\max\mathcal{A} = \sup A = 1\) .
E.g.,if \(\mathcal{A} = (0,1)\) ,then \(\sup\mathcal{A} = 1\) but \(\max\mathcal{A}\) does not exist!
观察 A. 5(集合的上确界与信道编码定理)在信息理论中,典型的信道编码定理表明一个(有限的)实数 \(\alpha\) 是集合 \(\mathcal{A}\) 的上确界。因此,为了证明这样的定理,必须证明 \(\alpha\) 同时满足上面的第 2 和第 3 属性,即,
正向/直接部分:\(\alpha {-} \epsilon\) 在 \(\mathcal{A}\) 中是可达的:\(({\forall}\varepsilon > 0)\left( {\exists}a_{0} {\in} \mathcal{A} \right)a_{0} > \alpha {-} \varepsilon\)(A.1.1)
和
反向部分:\(\underline{\alpha}\) 是 \(\mathcal{A}\) 中所有可达值的界限:\(\left( {\forall}a {\in} \mathcal{A} \right)a {\leq} \alpha\) 。(A.1.2)
Observation A. 5 (Supremum of a set and channel coding theorems) In information theory, a typical channel coding theorem establishes that a (finite) real number \(\alpha\) is the supremum of a set \(\mathcal{A}\) . Thus,to prove such a theorem,one must show that \(\alpha\) satisfies both properties 3 and 2 above,i.e.,
Forward/Direct part: \((\alpha {-} \epsilon)\) is achievable in \(\mathcal{A} : ({\forall}\varepsilon > 0)\left( {\exists}a_{0} {\in} \mathcal{A} \right)a_{0} > \alpha {-} \varepsilon\)
(A.1.1)
and
Converse part: \(\underline{\alpha}\) is a bound for all achievable values in \(\mathcal{A} : \left( {\forall}a {\in} \mathcal{A} \right)a {\leq} \alpha\) .
(A.1.2)
性质 A. 6(最大值的性质)
-
\(\left( {\forall}a {\in} \mathcal{A} \right)a {\leq} \max\mathcal{A}\) ,如果 \(\max\mathcal{A}\) 在 \(\mathbb{R} {\cup} \{{-} {\infty},{\infty}\}\) 中存在。
-
\(\max\mathcal{A} {\in} \mathcal{A}\) 。
- 根据上述性质,为了得到 \(\left\lbrack \alpha = \max\mathcal{A} \right\rbrack\),需要证明 \(\alpha\) 同时满足
反向部分:\(\underline{\alpha}\) 是 \(\mathcal{A}\) 中所有可达值的界限:\(\left( {\forall}a {\in} \mathcal{A} \right)a {\leq} \alpha\)。和
可达性/正向/直接部分:\(\underline{\alpha}\) 在 \(\mathcal{A}\) 中是可达的:\(\alpha {\in} \mathcal{A}\)
例如,计算二进制对称信道的信道容量。
Property A. 6 (Properties of the maximum)
-
\(\left( {\forall}a {\in} \mathcal{A} \right)a {\leq} \max\mathcal{A}\) ,if \(\max\mathcal{A}\) exists in \(\mathbb{R} {\cup} \{{-} {\infty},{\infty}\}\) .
-
\(\max\mathcal{A} {\in} \mathcal{A}\) .
- From the above property,in order to obtain \(\left\lbrack \alpha = \max\mathcal{A} \right\rbrack\) ,one needs to show that \(\alpha\) satisfies both
Converse part: \(\underline{\alpha}\) is a bound for all achievable values in \(\mathcal{A} : \left( {\forall}a {\in} \mathcal{A} \right)a {\leq} \alpha\) . and
Achievability/Forward/Direct part: \(\underline{\alpha}\) is achievable in \(\mathcal{A} : \alpha {\in} \mathcal{A}\)
E.g. Computation of the channel capacity of a binary symmetric channel.
A. 2 Infimum and minimum
下确界和最小值的概念与上确界和最大值的概念相对应。
The concepts of infimum and minimum are dual to those of supremum and maxi-mum.
定义 A. 7(集合的下界)如果实数 \({\ell}\) 是 \(\mathbb{R}\) 中非空子集 \(\mathcal{A}\) 的下界,即 \(\mathcal{A}\) 中的每个元素都大于或等于 \({\ell}\);我们说 \(\mathcal{A}\) 是有下界的。符号上,定义表示为:
Definition A. 7 (Lower bound of a set) A real number \({\ell}\) is called a lower bound of a non-empty subset \(\mathcal{A}\) in \(\mathbb{R}\) if every element of \(\mathcal{A}\) is greater than or equal to \({\ell}\) ; we say that \(\mathcal{A}\) is bounded below. Symbolically,the definition becomes:
\(\mathcal{A} {\subset} \mathbb{R}\) 有下界 \({\Leftrightarrow} \left( {\exists}{\ell} {\in} \mathbb{R} \right)\) 使得 \(\left( {\forall}a {\in} \mathcal{A} \right)a {\geq} {\ell}\)。
\(\mathcal{A} {\subset} \mathbb{R}\) is bounded below \({\Leftrightarrow} \left( {\exists}{\ell} {\in} \mathbb{R} \right)\) such that \(\left( {\forall}a {\in} \mathcal{A} \right)a {\geq} {\ell}\) .
定义 A. 8(最大下界或下确界)假设 \(\mathcal{A}\) 是 \(\mathbb{R}\) 的一个非空子集。如果实数 \({\ell}\) 是 \(\mathcal{A}\) 的下界,并且对于 \(\mathcal{A}\) 的任何其他下界 \({{\ell}}^{{\prime}}\),都有 \({\ell} {\geq} {{\ell}}^{{\prime}}\),那么我们说实数 \({\ell}\) 是 \(\mathcal{A}\) 的最大下界或下确界。在这种情况下,我们写 \({\ell} = \inf\mathcal{A}\);其他表示方法包括 \({\ell} = \mathop{\inf}\limits_{x {\in} \mathcal{A}}x\) 和 \({\ell} = \inf\{x : x {\in} \mathcal{A}\}\)。
Definition A. 8 (Greatest lower bound or infimum) Suppose \(\mathcal{A}\) is a non-empty subset of \(\mathbb{R}\) . Then we say that a real number \({\ell}\) is a greatest lower bound or infimum of \(\mathcal{A}\) if \({\ell}\) is a lower bound of \(\mathcal{A}\) and if \({\ell} {\geq} {{\ell}}^{{\prime}}\) for each lower bound \({{\ell}}^{{\prime}}\) of \(\mathcal{A}\) . In this case,we write \({\ell} = \inf\mathcal{A}\) ; other notations are \({\ell} = \mathop{\inf}\limits_{x {\in} \mathcal{A}}x\) and \({\ell} = \inf\{x : x {\in} \mathcal{A}\}\) .
完备性公理:(最大下界性质)让 \(\mathcal{A}\) 是一个有下界的非空子集 \(\mathbb{R}\)。那么 \(\mathcal{A}\) 存在最大下界。
Completeness Axiom: (Greatest lower bound property) Let \(\mathcal{A}\) be a
non-empty subset of \(\mathbb{R}\) that is bounded below. Then \(\mathcal{A}\) has a greatest lower bound.
A. 2 Infimum and minimum
-
It directly follows that if a non-empty set in \(\mathbb{R}\) has an infimum,then this infimum is unique.
-
By definition,the empty set \(({\varnothing})\) and any set not bounded below do not admit an infimum in \(\mathbb{R}\) .
-
直接推论,如果 \(\mathbb{R}\) 中的一个非空集合有下确界,那么这个下确界是唯一的。
-
根据定义,空集 \(({\varnothing})\) 和任何没有下界的集合在 \(\mathbb{R}\) 中都不存在下确界。
Property A. 10 (Properties of the infimum)
- 在 \(\mathbb{R} {\cup} \{{-} {\infty},{\infty}\}\) 中任何集合的下确界总是存在的。
这些扩展定义将在本课程中采用。
- The infimum of any set in \(\mathbb{R} {\cup} \{{-} {\infty},{\infty}\}\) always exists.
These extended definitions will be adopted in this course.
-
\(\left( {\forall}a {\in} \mathcal{A} {\subset} \mathbb{R}{\cup}\{{-} {\infty},{\infty}\} \right)a {\geq} \inf\mathcal{A}\)。
-
如果 \({\infty} > \inf\mathcal{A} > {-} {\infty}\),则对于任何 \(\varepsilon > 0\),都存在 \(a_{0} {\in} \mathcal{A}\) 使得 \(a_{0} < \inf\mathcal{A} + \varepsilon\)。
(在 \(\left| \inf\mathcal{A} \right| < {\infty}\) 的假设下,存在 \(a_{0} {\in} \lbrack\inf\mathcal{A},\inf\mathcal{A} + \varepsilon)\) 被称为下确界的逼近性质。)
(The existence of \(a_{0} {\in} \lbrack\inf\mathcal{A},\inf\mathcal{A} + \varepsilon)\) for any \(\varepsilon > 0\) under the assumption of \(\left| \inf\mathcal{A} \right| < {\infty}\) is called the approximation property for the infimum.)
-
如果 \(\inf\mathcal{A} = {-} {\infty}\),则对于任何 \(L {\in} \mathbb{R}\),都存在 \(B_{0} {\in} \mathcal{A}\) 满足 \(B_{0} < L\)。
-
如果 \(\inf\mathcal{A} = {\infty}\),则 \(\mathcal{A}\) 为空集。
定义 A. 9(最小值)如果 \(\inf\mathcal{A} {\in} \mathcal{A}\),则 \(\inf\mathcal{A}\) 也称为 \(\mathcal{A}\) 的最小值,并用 \(\min\mathcal{A}\) 表示。然而,如果 \(\inf\mathcal{A} {\notin} \mathcal{A}\),我们说 \(\mathcal{A}\) 的最小值不存在。
观察 A. 11(集合的下确界与信道编码定理)
Definition A. 9 (Minimum) If inf \(\mathcal{A} {\in} \mathcal{A}\) ,then inf \(\mathcal{A}\) is also called the min-imum of \(\mathcal{A}\) ,and is denoted by \(\min\mathcal{A}\) . However,if inf \(\mathcal{A} {\notin} \mathcal{A}\) ,we say that the minimum of \(\mathcal{A}\) does not exist.
Observation A. 11 (Infimum of a set and channel coding theorems)
在信息理论中,典型的源编码定理表明一个(有限的)实数 \(\alpha\) 是集合 \(\mathcal{A}\) 的下确界。因此,为了证明这样的定理,必须证明 \(\alpha\) 同时满足上面的第 2 和第 3 属性,即,
正向/直接部分:\((\alpha + \epsilon)\) 在 \(\mathcal{A}\) 中是可达的:\(({\forall}\varepsilon > 0)\left( {\exists}a_{0} {\in} \mathcal{A} \right)a_{0} < \alpha + \varepsilon\)
(A.2.1)
和
反向部分:\(\underline{\alpha}\) 是 \(\mathcal{A}\) 中所有可达值的界限:\(\left( {\forall}a {\in} \mathcal{A} \right)a {\geq} \alpha\)。
(A.2.2)
In information theory, a typical source coding theorem establishes that a (finite) real number \(\alpha\) is the infimum of a set \(\mathcal{A}\) . Thus,to prove such a theorem,one must show that \(\alpha\) satisfies both properties 3 and 2 above,i.e.,
Forward/Direct part: \((\alpha + \epsilon)\) is achievable in \(\mathcal{A} : ({\forall}\varepsilon > 0)\left( {\exists}a_{0} {\in} \mathcal{A} \right)a_{0} < \alpha + \varepsilon\)
(A.2.1)
and
Converse part: \(\underline{\alpha}\) is a bound for all achievable values in \(\mathcal{A} : \left( {\forall}a {\in} \mathcal{A} \right)a {\geq} \alpha\) .
(A.2.2)
性质 A. 12(最小值的性质)
-
\(\left( {\forall}a {\in} \mathcal{A} \right)a {\geq} \min\mathcal{A}\),如果 \(\min\mathcal{A}\) 在 \(\mathbb{R} {\cup} \{{-} {\infty},{\infty}\}\) 中存在。
-
\(\min\mathcal{A} {\in} \mathcal{A}\)。
- 根据上述性质,为了得到 \(\left\lbrack \alpha = \min\mathcal{A} \right\rbrack\),需要证明 \(\alpha\) 同时满足
反向部分:\(\underline{\alpha}\) 是 \(\mathcal{A}\) 中所有可达值的界限:\(\left( {\forall}a {\in} \mathcal{A} \right)a {\geq} \alpha\)。和
可达性/正向/直接部分:\(\underline{\alpha}\) 在 \(\mathcal{A}\) 中是可达的:\(\alpha {\in} \mathcal{A}\)
例如,计算二进制离散记忆无源(DMS)和汉明距离度量下的率失真函数(参见定理 6.23)。
Property A. 12 (Properties of the minimum)
-
\(\left( {\forall}a {\in} \mathcal{A} \right)a {\geq} \min\mathcal{A}\) ,if \(\min\mathcal{A}\) exists in \(\mathbb{R} {\cup} \{{-} {\infty},{\infty}\}\) .
-
\(\min\mathcal{A} {\in} \mathcal{A}\) .
- From the above property,in order to obtain \(\left\lbrack \alpha = \min\mathcal{A} \right\rbrack\) ,one needs to show that \(\alpha\) satisfies both
Converse part: \(\underline{\alpha}\) is a bound for all achievable values in \(\mathcal{A} : \left( {\forall}a {\in} \mathcal{A} \right)a {\geq} \alpha\) . and
Achievability/Forward/Direct part: \(\underline{\alpha}\) is achievable in \(\mathcal{A} : \alpha {\in} \mathcal{A}\)
E.g. Computation of the rate-distortion function for binary DMS and Hamming distance measure (cf. Theorem 6.23).
A. 3 Boundedness and suprema operations
定义 A. 13(有界性)如果 \(\mathbb{R}\) 的子集 \(\mathcal{A}\) 既有上界也有下界,我们称它为有界的;否则称为无界的。
Definition A. 13 (Boundedness) A subset \(\mathcal{A}\) of \(\mathbb{R}\) is said to be bounded if it is both bounded above and bounded below; otherwise it is called unbounded.
引理 A. 14(有界性的条件)一个 \(\mathbb{R}\) 的子集 \(\mathcal{A}\) 是有界的当且仅当存在实数 \(k\),使得集合中的所有元素 \(a\) 都满足 \(|a| \leq k\)。
Lemma A. 14 (Condition for boundedness) A subset \(\mathcal{A}\) of \(\mathbb{R}\) is bounded iff \(\left( {\exists}\ k {\in} \mathbb{R} \right)\) such that \(\left( {\forall}\ a {\in} \mathcal{A} \right)\ |a| {\leq} k.\)
引理 A.15(单调性质)假设 \(\mathcal{A}\) 和 \(\mathcal{B}\) 是非空的 \(\mathbb{R}\) 子集,并且 \(\mathcal{A} \subset \mathcal{B}\)。那么:
-
\(\sup\mathcal{A} \leq \sup\mathcal{B}\)。
-
\(\inf\mathcal{A} \geq \inf\mathcal{B}\)。
Lemma A.15 (Monotone property) Suppose that \(\mathcal{A}\) and \(\mathcal{B}\) are non-\(\textbf{empty subsets of}\mathbb{R}\) such that \(\mathcal{A} {\subset} \mathcal{B}\) . Then
-
\(\sup\mathcal{A} {\leq} \sup\mathcal{B}\) .
-
\(\inf\mathcal{A} {\geq} \inf\mathcal{B}\) .
引理 A.16(集合运算的上确界)定义两个集合 \(\mathcal{A}\) 和 \(\mathcal{B}\) 的“加法”为:
定义一个集合 \(\mathcal{A}\) 与一个标量 \(k \in \mathbb{R}\) 的“标量乘法”为:
最后,定义集合 \(\mathcal{A}\) 的“取反”为:
那么以下结论成立:
-
如果 \(\mathcal{A}\) 和 \(\mathcal{B}\) 都有上界,那么 \(\mathcal{A} + \mathcal{B}\) 也有上界,并且 \(\sup(\mathcal{A} + \mathcal{B}) = \sup\mathcal{A} + \sup\mathcal{B}\)。
-
如果 \(0 < k < \infty\) 并且 \(\mathcal{A}\) 有上界,那么 \(k \cdot \mathcal{A}\) 也有上界,并且 \(\sup(k \cdot \mathcal{A}) = k \cdot \sup\mathcal{A}\)。
-
\(\sup\mathcal{A} = {-} \inf({-}\mathcal{A})\) 并且 \(\inf\mathcal{A} = {-} \sup({-}\mathcal{A})\)。
-
If \(\mathcal{A}\) and \(\mathcal{B}\) are both bounded above,then \(\mathcal{A} + \mathcal{B}\) is also bounded above and \(\sup\left( \mathcal{A} + \mathcal{B} \right) = \sup\mathcal{A} + \sup\mathcal{B}.\)
-
If \(0 < k < {\infty}\) and \(\mathcal{A}\) is bounded above,then \(k {\cdot} \mathcal{A}\) is also bounded above and \(\sup\left( k {\cdot} \mathcal{A} \right) = k {\cdot} \sup\mathcal{A}\) .
-
\(\sup\mathcal{A} = {-} \inf\left( {-}\mathcal{A} \right)\quad\) and \(\quad\inf\mathcal{A} = {-} \sup\left( {-}\mathcal{A} \right)\) .
A. 3 Boundedness and suprema operations
- 对于两个集合的“乘积”,性质 1 不成立,其中集合 \(\mathcal{A}\) 和 \(\mathcal{B}\) 的“乘积”定义为:
在这种情况下,可能发生以下两种情况:
例如,\(\mathcal{A} = \lbrack {-} 1,0)\) 和 \(\mathcal{B} = \lbrack {-} 1,0)\) 。则
例如,\(\mathcal{A} = \lbrack {-} 1,0)\) 和 \(\mathcal{B} = \lbrack 0,1)\) 。则
A. 3 Boundedness and suprema operations
引理 A. 17(单调函数的上确界/下确界)
- 如果 \(f : \mathbb{R} \rightarrow \mathbb{R}\) 是一个非递减函数,则有:
并且
- 如果 \(f : \mathbb{R} \rightarrow \mathbb{R}\) 是一个非递增函数,则有:
并且
Lemma A. 17 (Supremum/infimum for monotone functions)
- If \(f : \mathbb{R} {\rightarrow} \mathbb{R}\) is a non-decreasing function,then
and
- If \(f : \mathbb{R} {\rightarrow} \mathbb{R}\) is a non-increasing function,then
and
Illustration of Lemma A. 17

-
\(\mathbb{N}\) 表示“自然数”(正整数)\(1,2,3,{\cdots}\)。
-
从实值函数中得到的序列表示为:
换句话说,\(f(n)\) 对每个 \(n = 1,2,3,{\ldots}\) 都是一个实数。
- 通常我们写 \(f(n) = a_n\),并且我们经常用以下任一符号表示序列:
- 关于序列的一个重要问题是,当 \(n\) 变大时会发生什么。具体来说,我们想知道当 \(n\) 足够大时,每个 \(a_n\) 是否接近某个固定数 \(L\)(即 \(\left. a_n \right)\) 的极限)。
定义 A. 18(极限)序列 \({\left\{a_n \right\}}_{n = 1}^{\infty}\) 的极限是实数 \(L\),满足:\(({\forall}\varepsilon > 0)({\exists}N)\) 使得 \(({\forall}n > N)\)
在这种情况下,我们写 \(L = \mathop{\lim}\limits_{n \rightarrow \infty}a_n\)。如果没有这样的 \(L\) 满足上述条件,我们说序列 \({\left\{a_n \right\}}_{n = 1}^{\infty}\) 的极限不存在。
性质 A.19 如果序列 \({\left\{a_n \right\}}_{n = 1}^{\infty}\) 和 \({\left\{b_n \right\}}_{n = 1}^{\infty}\) 都在 \(\mathbb{R}\) 中有极限,则以下成立:
- \(\mathop{\lim}\limits_{n \rightarrow \infty}\left( a_n + b_n \right) = \mathop{\lim}\limits_{n \rightarrow \infty}a_n + \mathop{\lim}\limits_{n \rightarrow \infty}b_n\)。
- \(\mathop{\lim}\limits_{n \rightarrow \infty}\left( \alpha \cdot a_n \right) = \alpha \cdot \mathop{\lim}\limits_{n \rightarrow \infty}a_n\)。
- \(\mathop{\lim}\limits_{n {\rightarrow} {\infty}}\left( a_{n}b_{n} \right) = \left( \mathop{\lim}\limits_{n {\rightarrow} {\infty}}a_{n} \right)\left( \mathop{\lim}\limits_{n {\rightarrow} {\infty}}b_{n} \right)\) .
Definition A. 18 (Limit) The limit of \({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\) is the real number \(L\) satisfying: \(({\forall}\varepsilon > 0)({\exists}N)\) such that \(({\forall}n > N)\)
In this case,we write \(L = \mathop{\lim}\limits_{n {\rightarrow} {\infty}}a_{n}\) . If no such \(L\) satisfies the above statement, we say that the limit of \({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\) does not exist.
Property A.19 If \({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\) and \({\left\{b_{n} \right\}}_{n = 1}^{{\infty}}\) both have a limit in \(\mathbb{R}\) ,then the fol-lowing hold.
-
\(\mathop{\lim}\limits_{n {\rightarrow} {\infty}}\left( a_{n} + b_{n} \right) = \mathop{\lim}\limits_{n {\rightarrow} {\infty}}a_{n} + \mathop{\lim}\limits_{n {\rightarrow} {\infty}}b_{n}\) .
-
\(\mathop{\lim}\limits_{n {\rightarrow} {\infty}}\left( \alpha {\cdot} a_{n} \right) = \alpha {\cdot} \mathop{\lim}\limits_{n {\rightarrow} {\infty}}a_{n}\) .
-
\(\mathop{\lim}\limits_{n {\rightarrow} {\infty}}\left( a_{n}b_{n} \right) = \left( \mathop{\lim}\limits_{n {\rightarrow} {\infty}}a_{n} \right)\left( \mathop{\lim}\limits_{n {\rightarrow} {\infty}}b_{n} \right)\) .
A. 4 Sequences and their limits
- 请注意,在上述定义中,\({-} {\infty}\) 和 \({\infty}\) 不能作为任何序列的合法极限。
Note that in the above definition, \({-} {\infty}\) and \({\infty}\) cannot be a legitimate limit for any sequence.
-
实际上,如果对于所有的 \(L\) 都存在 \(N\) 使得当 \(n > N\) 时,\(a_{n} > L\),那么我们说 \(a_{n}\) 发散到 \({\infty}\),并写作 \(a_{n} {\rightarrow} {\infty}\)。对于 \(a_{n}\) 发散到 \({-} {\infty}\) 也有类似的说法。
-
为了方便,我们将在扩展的实数集中工作,因此声明发散到 \({\infty}\) 或 \({-} {\infty}\) 的序列 \({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\) 在 \(\mathbb{R} {\cup} \{{-} {\infty},{\infty}\}\) 中有极限。
引理 A.20(单调序列的收敛性)如果 \({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\) 随 \(n\) 增加而非递减,则 \(\mathop{\lim}\limits_{n {\rightarrow} {\infty}}a_{n}\) 在 \(\mathbb{R} {\cup} \{{-} {\infty},{\infty}\}\) 中存在。如果 \({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\) 还被限制在上界 - 即对于某些实数 \(L\),所有的 \(n\) 都有 \(a_{n} {\leq} L\),那么 \(\mathop{\lim}\limits_{n {\rightarrow} {\infty}}a_{n}\) 在 \(\mathbb{R}\) 中存在。
同样地,如果 \({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\) 随 \(n\) 增加而非递增,则 \(\mathop{\lim}\limits_{n {\rightarrow} {\infty}}a_{n}\) 在 \(\mathbb{R} {\cup} \{{-} {\infty},{\infty}\}\) 中存在。如果 \({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\) 还被限制在下界 - 即对于某些实数 \(L\),所有的 \(n\) 都有 \(a_{n} {\geq} L\),那么 \(\mathop{\lim}\limits_{n {\rightarrow} {\infty}}a_{n}\) 在 \(\mathbb{R}\) 中存在。
Lemma A.20 (Convergence of monotone sequences) If \({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\) is non-decreasing in \(n\) ,then \(\mathop{\lim}\limits_{n {\rightarrow} {\infty}}a_{n}\) exists in \(\mathbb{R} {\cup} \{{-} {\infty},{\infty}\}\) . If \({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\) is also bounded from above - i.e., \(a_{n} {\leq} L{\forall}n\) for some \(L\) in \(\mathbb{R} {-}\) then \(\mathop{\lim}\limits_{n {\rightarrow} {\infty}}a_{n}\) exists in \(\mathbb{R}\) .
Likewise,if \({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\) is non-increasing in \(n\) ,then \(\mathop{\lim}\limits_{n {\rightarrow} {\infty}}a_{n}\) exists in \(\mathbb{R} {\cup} \{{-} {\infty},{\infty}\}\) . If \({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\) is also bounded from below - i.e., \(a_{n} {\geq} L{\forall}n\) for some \(L\) in \(\mathbb{R} {-}\) then \(\mathop{\lim}\limits_{n {\rightarrow} {\infty}}a_{n}\) exists in \(\mathbb{R}.\)
A. 4 Sequences and their limits
- 序列的极限可能不存在。
例如,\(a_{n} = {(-1)}^{n}\)。
那么对于较大的 \(n\),\(a_{n}\) 将接近 -1 或 1。
- 因此,需要更通用的定义来描述序列的一般极限行为。
定义 A. 21(上极限和下极限)\({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\) 的上极限是在 \(\mathbb{R} {\cup} \{{-} {\infty},{\infty}\}\) 中定义的扩展实数
而 \({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\) 的下极限是通过以下方式定义的扩展实数
有些人还使用符号 \(\overline{\lim}\) 和 \(\underline{\lim}\) 分别表示上极限和下极限。
- Hence, more generalized definitions that can describe the general limiting be-havior of a sequence is required.
Definition A. 21 (limsup and liminf) The limit supremum of \({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\) is the extended real number in \(\mathbb{R} {\cup} \{{-} {\infty},{\infty}\}\) defined by
and the limit infimum of \({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\) is the extended real number defined by
Some also use the notations \(\overline{\lim}\) and \(\underline{\lim}\) to denote limsup and liminf,respectively.
- 注意,序列的上极限和下极限总是在 \(\mathbb{R} {\cup} \{{-} {\infty},{\infty}\}\) 中定义,因为序列 \(\mathop{\sup}\limits_{k {\geq} n}a_{k} = \sup\left\{a_{k} : k {\geq} n \right\}\) 和 \(\mathop{\inf}\limits_{k {\geq} n}a_{k} = \inf\left\{a_{k} : k {\geq} n \right\}\) 随 \(n\) 单调(参见引理 A.20)。
引理 A.22(极限)对于序列 \({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\),
- Note that the limit supremum and the limit infimum of a sequence is always defined in \(\mathbb{R} {\cup} \{{-} {\infty},{\infty}\}\) ,since the sequences \(\mathop{\sup}\limits_{k {\geq} n}a_{k} = \sup\left\{a_{k} : k {\geq} n \right\}\) and \(\mathop{\inf}\limits_{k {\geq} n}a_{k} = \inf\left\{a_{k} : k {\geq} n \right\}\) are monotone in \(n\) (cf. Lemma A.20).
Lemma A.22 (Limit) For a sequence \({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\) ,
Property A. 23 (Properties of the limit supremum)
- 上极限在 \(\mathbb{R} {\cup} \{{-} {\infty},{\infty}\}\) 中总是存在的。
The limit supremum always exists in \(\mathbb{R} {\cup} \{{-} {\infty},{\infty}\}\) .
- 如果 \(\left| \limsup_{m \rightarrow {\infty}} a_{m} \right| < {\infty}\),则 \(({\forall}\varepsilon > 0)({\exists}N)\) 使得 \(({\forall}n > N)a_{n} < \lim\mathop{\sup}\limits_{m {\rightarrow} {\infty}}a_{m} + \varepsilon\)。 (注意这适用于所有 \(n > N\)。)
If \(\left| \limsup a_{m {\rightarrow} {\infty}}a_{m} \right| < {\infty}\) ,then \(({\forall}\varepsilon > 0)({\exists}N)\) such that \(({\forall}n > N)a_{n} <\) \(\lim\mathop{\sup}\limits_{m {\rightarrow} {\infty}}a_{m} + \varepsilon\) . (Note that this holds for every \(n > N\) .)
- 如果 \(\left| \limsup_{m \rightarrow {\infty}} a_{m} \right| < {\infty}\),则 \(\left( {\forall}\varepsilon > 0\text{且为整数}K \right)({\exists}N > K)\) 使得 \(a_{N} > \limsup_{m \rightarrow {\infty}}a_{m} - \varepsilon\)。 (注意这只适用于一个大于 \(K\) 的 \(N\)。)
If \(\left| \limsup{}_{m {\rightarrow} {\infty}}a_{m} \right| < {\infty}\) ,then \(\left( {\forall}\varepsilon > 0\text{and integer}K \right)({\exists}N > K)\) such that \(a_{N} > \limsup{}_{m {\rightarrow} {\infty}}a_{m} {-} \varepsilon\) . (Note that this holds only for one \(N\) ,which is larger than \(K\) .)
Property A. 24 (Properties of the limit infimum)
- 下极限在 \(\mathbb{R} {\cup} \{{-} {\infty},{\infty}\}\) 中总是存在的。
The limit infimum always exists in \(\mathbb{R} {\cup} \{{-} {\infty},{\infty}\}\) .
-
如果 \(\left| \mathop{\liminf}\limits_{m {\rightarrow} {\infty}}a_{m} \right| < {\infty}\),则 \(\left( {\forall}\varepsilon > 0\text{和}K \right)({\exists}N > K)\) 使得 \(a_{N} < \mathop{\liminf}\limits_{m {\rightarrow} {\infty}}a_{m} + \varepsilon\)。 (注意这只适用于一个大于 \(K\) 的 \(N\)。)
-
如果 \(\left| \mathop{\liminf}\limits_{m {\rightarrow} {\infty}}a_{m} \right| < {\infty}\),则 \(({\forall}\varepsilon > 0)({\exists}N)\) 使得 \(({\forall}n > N)a_{n} > \liminf_{m {\rightarrow} {\infty}}a_{m} - \varepsilon\)。 (注意这适用于所有 \(n > N\)。)
定义 A. 25(足够大)如果一个性质对于序列 \({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\) 几乎总是成立,或者对所有足够大的 \(n\) 成立,那么我们说这个性质在某个 \(N\) 之后的每个 \(n > N\) 都成立。
定义 A. 26(无限次数)如果一个性质对于序列 \({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\) 无限次成立,或者对无限多的 \(n\) 成立,那么我们说对于每个 \(K\),这个性质在某个特定的 \(N > K\) 上成立。
Definition A. 25 (Sufficiently large) We say that a property holds for a se-quence \({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\) almost always or for all sufficiently large \(n\) if the property holds for every \(n > N\) for some \(N\) .
Definition A. 26 (Infinitely often) We say that a property holds for a sc-quence \({\left\{a_{n} \right\}}_{n = 1}^{{\infty}}\) infinitely often or for infinitely many \(n\) if for every \(K\) ,the property holds for one (specific) \(N\) with \(N > K\) .
- 那么,性质 A. 23 的第 2 和第 3 条可以分别重新表述为:
如果 \(\left| \lim\mathop{\sup}\limits_{m {\rightarrow} {\infty}}a_{m} \right| < {\infty}\),则 \(({\forall}\varepsilon > 0)\)
和
- 类似地,性质 A.24 的第 2 和第 3 条变为:如果 \(\left| \mathop{\liminf}\limits_{m {\rightarrow} {\infty}}a_{m} \right| < {\infty}\),则 \(({\forall}\varepsilon > 0)\)
和
- Then Properties 2 and 3 of Property A. 23 can be respectively re-phrased as:
if \(\left| \lim\mathop{\sup}\limits_{m {\rightarrow} {\infty}}a_{m} \right| < {\infty}\) ,then \(({\forall}\varepsilon > 0)\)
and
- Similarly,Properties 2 and 3 of Property A.24 becomes: if \(\left| \mathop{\liminf}\limits_{m {\rightarrow} {\infty}}a_{m} \right| <\) \({\infty}\) ,then \(({\forall}\varepsilon > 0)\)
and
A. 4 Sequences and their limits
引理 A.27:
-
\(\mathop{\liminf}\limits_{n \rightarrow \infty}a_{n} \leq \lim\mathop{\sup}\limits_{n \rightarrow \infty}a_{n}\)。
-
如果对于所有足够大的 \(n\),\(a_{n} \leq b_{n}\),那么
-
\(\lim\mathop{\sup}\limits_{n \rightarrow \infty}a_{n} < r \Rightarrow a_{n} < r\) 对所有足够大的 \(n\) 成立。
-
\(\lim\mathop{\sup}\limits_{n \rightarrow \infty}a_{n} > r \Rightarrow a_{n} > r\) 对无限多的 \(n\) 成立。
- 如果 \(\mathop{\lim}\limits_{n \rightarrow \infty}a_{n}\) 存在,则
且
-
上极限等于最大的聚点。
-
下极限等于最小的聚点。
-
聚点是序列 \(a_{n}\) 无限次接近的点。
-
Limsup \(=\) largest clustering point
-
Liminf \(=\) smallest clustering point
-
- A clustering point is a point that the sequence \(a_{n}\) hits close for infinitely many times.
例如,\(a_{n} = \sin(n\pi/2)\)
\({\Rightarrow} {\left\{a_{n} \right\}}_{n \geq 1} = \{1,0, -1,0,1,0, -1,0,{\ldots}\}\)
这个序列中有三个聚点,分别是 \(-1, 0\) 和 \(1\)。
因此,
例如,\(a_{n} = -n\)。那么 \(\lim\mathop{\sup}\limits_{n \rightarrow \infty}a_{n} = \mathop{\liminf}\limits_{n \rightarrow \infty}a_{n} = -\infty\)。
例如,\(a_{n} = n\)。那么 \(\lim\mathop{\sup}\limits_{n \rightarrow \infty}a_{n} = \mathop{\liminf}\limits_{n \rightarrow \infty}a_{n} = \infty\)
。
A. 5 Equivalence
-
我们通过提供一些常用于简化证明的等价陈述来结束这个附录。
-
例如,不是直接证明量 \(x\) 小于等于量 \(y\),可以取一个任意常数 \(\varepsilon > 0\) 并证明 \(x < y + \varepsilon\)。
-
由于 \(y + \varepsilon\) 是比 \(y\) 大的量,在某些情况下,证明 \(x < y + \varepsilon\) 可能比证明 \(x \leq y\) 更容易。定理 A. 28 对于任何在 \(\mathbb{R}\) 中的 \(x, y\) 和 \(a\),
-
对所有 \(\varepsilon > 0\),\(x < y + \varepsilon\) 当且仅当 \(x \leq y\);
-
对某个 \(\varepsilon > 0\),\(x < y - \varepsilon\) 当且仅当 \(x < y\);
-
对所有 \(\varepsilon > 0\),\(x > y - \varepsilon\) 当且仅当 \(x \geq y\);
-
对某个 \(\varepsilon > 0\),\(x > y + \varepsilon\) 当且仅当 \(x > y\);
-
对所有 \(\varepsilon > 0\),\(|a| < \varepsilon\) 当且仅当 \(a = 0\)。
key notes
-
实数线上子集的上确界和下确界
-
上极限和下极限(及其性质)
-
足够大和无限次
-
等价性

浙公网安备 33010602011771号