假设 C ⊂ R n C \subset \mathbb{R}^n C⊂Rn 是凸集。设 x i ∈ C x_i \in C xi∈C 且 θ i ≥ 0 , ∑ i = 1 ∞ θ i = 1 \theta_i \geq 0,\sum_{i=1}^\infty{\theta_i}=1 θi≥0,∑i=1∞θi=1。证明如果 x = ∑ i = 1 ∞ θ i x i x = \sum_{i=1}^\infty{\theta_ix_i} x=∑i=1∞θixi收敛,那么 x ∈ C x\in C x∈C。
证:
1.对扩展到无穷和的合理性说明:
凸集对于有限个点的定义表明,对于任意两个点 x , y ∈ C x,y∈C x,y∈C以及任意实数α满足 0 ≤ α ≤ 1 0≤α≤1 0≤α≤1,有 α x + ( 1 − α ) y ∈ C α x+(1−α)y∈C αx+(1−α)y∈C。对于有限个点 x 1 , x 2 , ⋯ , x k ∈ C x_{1},x_{2},\cdots,x_{k}∈C x1,x2,⋯,xk∈C 以及非负的 θ 1 , θ 2 , ⋯ , θ k θ_{1},θ_{2},\cdots,θ_{k} θ1,θ2,⋯,θk且 ∑ i = 1 k θ i = 1 \sum_{i=1}^{k} θ_{i}=1 ∑i=1kθi=1,
那么 ∑ i = 1 k θ i x i ∈ C \sum_{i=1}^{k} θ_{i}x_{i} \in C ∑i=1kθixi∈C。现在考虑无穷和的情况。由于 ∑ i = 1 ∞ θ i x i \sum_{i=1}^∞ θ_{i}x_{i} ∑i=1∞θixi收敛,对于任意给定的 ε > 0 ε>0 ε>0,存在一个整数N,使得对于所有的 m , n > N m,n>N m,n>N,有 ∥ ∑ i = 1 m θ i x i − ∑ i = 1 n θ i x i ∥ < ε \left \| \sum_{i=1}^ mθ_{i}x_{i}-\sum_{i=1}^ nθ_{i}x_{i}\right \| <ε ∥∑i=1mθixi−∑i=1nθixi∥<ε。这表明部分和序列是一个柯西序列。在 R m \mathbb{R}^m Rm 中,柯西序列收敛。设 s n = ∑ i = 1 n θ i x i s_{n}=\sum_{i=1}^ nθ_{i} x_{i} sn=∑i=1nθixi。因为每个 s n s_{n} sn 都是 C C C 中各点的凸组合,所以对于所有的n, s n s_{n} sn都属于C。我们可以认为极限 x = lim n → ∞ s n x=\lim_{n \to \infty} s_{n} x=limn→∞sn可以被C中的元素任意逼近。
- 对凸集的闭性的澄清:
在像 R n \mathbb{R}^{n} Rn这样的有限维空间中,如果一个凸集包含它的所有极限点,那么它就是闭集。这是因为在有限维赋范空间中,一个集合是闭集当且仅当它是完备的。由于 R n \mathbb{R}^{n} Rn是完备的,并且凸集 C ⊂ R n C \subset \mathbb{R}^{n} C⊂Rn是由一组线性不等式定义的,所以可以证明C是闭半空间的交集,因此是闭集。
- 更严格地处理收敛性:
因为我们已经证明了部分和序列 s n {s_{n}} sn收敛到 x x x,并且每个 s n ∈ C s_{n} \in C sn∈C。现在,由于 C C C是闭集,它包含它的所有极限点。因此, x x x作为序列 s n {s_{n}} sn的极限,必然属于 C C C。
结论:
综上,可以得出结论,如果 x = ∑ i = 1 ∞ θ i x i x=\sum_{i=1}^{\infty}\limits \theta_ix_i x=i=1∑∞θixi在集合 C C C中收敛,那么 x x x一定属于 C C C。
解题思路:
设定与前提:
假设集合 C C C是实数集 R n \mathbb{R}^n Rn的一个凸集。
给定任意 x i x_i xi属于C,且存在 θ i ≥ 0 θ_i≥0 θi≥0,满足 ∑ θ i = 1 ∑θ_i = 1 ∑θi=1。
考虑凸集的性质:
凸集C的定义是:对于 C C C中的任意两点x和y,以及任意实数 α α α( 0 ≤ α ≤ 1 0≤α≤1 0≤α≤1),都有 α x + ( 1 − α ) y αx+(1-α)y αx+(1−α)y也属于 C C C。
由此,我们可以推断出,对于C中的任意有限个点 x i x_i xi和满足 ∑ θ i = 1 ∑θ_i = 1 ∑θi=1的非负实数 θ i θ_i θi,点 x = ∑ θ i ∗ x i x = ∑θ_i * x_i x=∑θi∗xi也必然属于 C C C。
收敛性的分析:
假设序列 { x n } \{x_n\} {xn}= { ∑ θ i ∗ x i n } \{∑θ_i * x_i^{n}\} {∑θi∗xin}(其中 x i n x_i^{n} xin是C中的点,且 θ i θ_i θi满足上述条件)收敛于 x x x。
由于每个 x i n x_i^n xin都属于 C C C,且 C C C是凸集,因此每个 x n x_n xn也都属于 C C C。
利用凸集的闭包性质:
凸集 C C C不仅是凸的,还是闭的(在有限维空间中,凸集总是闭的)。
这意味着,如果序列 { x n } \{x_n\} {xn}在 C C C中收敛,那么其极限点 x x x也必然属于 C C C。
得出结论:
综合以上分析,我们可以得出结论:如果 x = ∑ θ i ∗ x i x = ∑θ_i * x_i x=∑θi∗xi在集合 C C C中收敛,那么 x x x必然属于 C C C。
分析:
这个证明的关键在于理解凸集的性质,特别是凸集的闭包性质。由于凸集在有限维空间中总是闭的,因此我们可以利用这一性质来推断收敛序列的极限点也属于凸集。此外,凸集的定义也为我们提供了判断点是否属于凸集的有效方法。
英文原始内容:
Suppose C ⊂ R n C \subset \mathbb{R}^n C⊂Rn is convex. Let x i ∈ C x_i \in C xi∈C and θ i ≥ 0 , ∑ i = 1 ∞ θ i = 1 \theta_i \geq 0, \sum_{i=1}^\infty \theta_i = 1 θi≥0,∑i=1∞θi=1. Show that if x = ∑ i = 1 ∞ θ i x i x = \sum_{i=1}^\infty \theta_ix_i x=∑i=1∞θixi converges, then x ∈ C x \in C x∈C.
Proof:
Let C ⊂ R n C \subset \mathbb{R}^n C⊂Rn be a convex set. Let x i ∈ C x_i \in C xi∈C and θ i ≥ 0 \theta_i\geq0 θi≥0, such that ∑ i = 1 ∞ θ i = 1 \sum_{i=1}^\infty \theta_i = 1 ∑i=1∞θi=1.
- Justification for extending to infinite sums:
The definition of convexity for a finite number of points states that for any two points x , y ∈ C x, y \in C x,y∈C and any real number α \alpha α with 0 ≤ α ≤ 1 0\leq\alpha\leq1 0≤α≤1, we have α x + ( 1 − α ) y ∈ C \alpha x + (1-\alpha)y \in C αx+(1−α)y∈C. For a finite set of points x 1 , x 2 , … , x k ∈ C x_1,x_2,\ldots,x_k \in C x1,x2,…,xk∈C and non-negative θ 1 , θ 2 , … , θ k \theta_1,\theta_2,\ldots,\theta_k θ1,θ2,…,θk with ∑ i = 1 k θ i = 1 \sum_{i=1}^k \theta_i = 1 ∑i=1kθi=1, it follows that ∑ i = 1 k θ i x i ∈ C \sum_{i=1}^k \theta_ix_i \in C ∑i=1kθixi∈C. Now, consider the infinite sum case. As the sum ∑ i = 1 ∞ θ i x i \sum_{i=1}^{\infty} \theta_i x_i ∑i=1∞θixi converges, for any given ε > 0 \varepsilon > 0 ε>0, there exists an integer N N N such that for all m , n > N , ∥ ∑ i = 1 n θ i x i − ∑ i = 1 m θ i x i ∥ < ε m, n>N, \left\|\sum_{i=1}^{n} \theta_i x_i-\sum_{i=1}^{m} \theta_i x_i\right\|<\varepsilon m,n>N,∥∑i=1nθixi−∑i=1mθixi∥<ε. This shows that the sequence of partial sums is a Cauchy sequence. In R n \mathbb{R}^n Rn, Cauchy sequences converge. Let s n = ∑ i = 1 n θ i x i s_n=\sum_{i=1}^{n} \theta_i x_i sn=∑i=1nθixi. Since each s n s_n sn is a convex combination of points in C C C and hence belongs to C C C for all n, we can think of the limit x = lim n → ∞ s n x=\lim_{n \to \infty} s_n x=limn→∞sn as being approximated arbitrarily closely by elements in C C C.
- Clarification on closure of convex sets:
In finite-dimensional spaces like R n \mathbb{R}^n Rn , a convex set is closed if it contains all its limit points. This follows from the fact that in a finite-dimensional normed space, a set is closed if and only if it is complete. Since R n \mathbb{R}^n Rn is complete and a convex set C ⊂ R n C \subset \mathbb{R}^n C⊂Rn is defined by a collection of linear inequalities, it can be shown that C is an intersection of closed half-spaces, and hence is closed.
- Handling convergence more rigorously:As we have shown that the sequence of partial sums s n {s_n} sn converges to x, and each s n ∈ C s_n\in C sn∈C. Now, since C is closed, it contains all its limit points. Therefore, x, being the limit of the sequence { s n } \left\{s_n\right\} {sn}, must belong to C.
Conclusion:In summary, we can conclude that if x= ∑ i = 1 ∞ a i x i \sum_{i=1}^{\infty}a_ix_i ∑i=1∞aixi converges in the set C, then x must belong to C.