Answers:
在通过常规程序简化了问题之后,可以通过将其转换为对偶最小化程序来解决该问题,该程序具有众所周知的答案并带有基本证明。也许这种二元化是问题中提到的“微妙步骤”。不平等也可以在纯机械方式通过最大化建立 通过拉格朗日乘数。
不过,首先,我根据最小二乘的几何形状提供了一种更优雅的解决方案。它不需要初步的简化,几乎是立即的,可以直接直观地了解结果。正如问题中所建议的那样,该问题简化为柯西-舒瓦兹不等式。
将视为具有通常点积的欧几里得空间中的n维向量。让ÿ = (0 ,0 ,... ,0 ,1 ,0 ,... ,X和ÿ为的正交投影X和ÿ到的正交补1是我个基础矢量和 1 = (1 ,1 ,... ,1 )。写。(在统计学术语,它们是相对于该装置的残差。)然后,由于和小号= | | X | | / √,
是的部件Ÿ在X方向。由柯西-施瓦茨,它被精确地最大化时X是平行于Ý = (- 1 ,- 1 ,... ,- 1 ,ñ - 1 ,- 1 ,- 1 ,... ,- 1 )/ Ñ,为此Ť 我 = ± √QED。
顺便说一句,这个解决方案提供的所有情况下的详尽描述最大化:它们都是形式
对于所有实数。
该分析很容易地概括为被任何一组回归变量代替的情况。显然,最大的Ť 我正比于剩余的长度ÿ,| | ÿ | | 。
由于在位置和尺度的变化下是不变的,因此我们可以不失一般性地假设X i为零,并且它们的平方为n - 1。这标识| Ť 我| 与| X i | ,因为S(均方根)是1。最大化等于最大化| Ť 我| 2 = T 2 我。取 i =不会失去一般性,或者说,由于 X 我是可更换的。
对偶问题是固定的值,并询问剩余X j,j ≠ 1的值是多少,以使平方和∑ n j = 1 X 2 j最小,假设∑ n j = 1 X j = 0。因为给定了X 1,所以这是使∑ n j = 2 X 2最小的问题。鉴于Σ Ñ。
该解决方案可以通过多种方式轻松找到。最基本的方法之一就是写
对于其中。扩展目标函数并使用此零和恒等式简化它会产生
立即表现出特有的解决方案是对所有Ĵ。对于此解决方案,
和
QED.
Return to the simplified program we began with:
subject to
The method of Lagrange multipliers (which is almost purely mechanical and straightforward) equates a nontrivial linear combination of the gradients of these three functions to zero:
Component by component, these equations are
The last of them imply either or . (We may rule out the latter case because then the first equation implies , trivializing the linear combination.) The sum-to-zero constraint produces . The sum-of-squares constraint provides the two solutions
They both yield
The inequality as stated is true. It is quite clear intuitively that we get the most difficult case for the inequality (that is, maximizing the left hannd side for given ) by choosing one value, say as large as possible while having all the others equal. Let us look at an example with such configuration:
EDIT
We will now prove the claim, as hinted above. First, for any given vector in this problem, we can replace it with without changing either side of the inequality above. So, in the following let us assume that . We can also by relabelling assume that is largest. Then, by choosing first and then we can check by simple algebra that we have equality in the claimed inequality. So, it is sharp.
Then, define the (convex) region by