珍珠湾ART

标题: 保 险 箱 和 钥 匙 [打印本页]

作者: 野 菜 花    时间: 2005-4-13 22:45
标题: 保 险 箱 和 钥 匙

10 个 保 险 箱 , 10把 钥 匙 , 每 把 钥 匙 恰 好 能 开 一 个 保 险 箱 , 每 个 保 险 箱 也 只 有 一 把 钥 匙 能 打 开 。 钥 匙 锁 在 保 险 箱 里 , 每 个 保 险 箱 里 一 把 钥 匙 , 方 法 是 随 机 的 。 先 打 破 3 个 箱 子 取 出 钥 匙 。 问 其 余 的 保 险 箱 均 能 用 钥 匙 打 开 的 概 率 是 多 少 ?
www.ddhw.com

 

作者: fzy    时间: 2005-4-14 02:33
标题: 回复:保 险 箱 和 钥 匙

The problem looks very hard.  I have only this silly method, and have not yet got the number.
 
Let P(n,k) be the probability of open all n safes with k keys. We need P(10,3).
P(n,1) = 1/n.www.ddhw.com
P(n,2) = [2*(n-2)/C(n,2)]*P(n-2,1) + [C(n-2,2)/C(n,2)]*P(n-2,2)
P(n,3) = [3*(n-3)/C(n,3)]*P(n-3,1) + [3*C(n-3,2)/C(n,3)]*P(n-3,2) + [C(n-3,3)/C(n,3)]*P(n-3,3)
 
From these recursive relations we should be able to get P(10,3). Need to write a program to calculate the number.
 
www.ddhw.com

 

作者: 乱弹    时间: 2005-4-14 02:56
标题: 真是非常难。我觉得跟置换群有关。

能开的情况是最多有三个轮换,并且每个轮换里都有一个被抓出来。这样是能直接计算出来的。可惜我对这些组合关系不熟悉,要做的话,又要查书,几乎每步都要仔细验证,非一时半会可以做好。www.ddhw.com
 
也许野菜花有简单办法吧。www.ddhw.com
 
不管怎样,这题目实在有意思。
www.ddhw.com

 

作者: 野 菜 花    时间: 2005-4-14 05:00
标题: 回复:回复:保 险 箱 和 钥 匙

In the original problem, in stead of 10 and 3,  n and k are used just as you did in your solution.
In the solution i know, comjecture the answer first (the answer looks very simple), then use the induction.
Don't hurry, take your time, otherwise you guys would feel bored again.
www.ddhw.com

 

作者: 乱弹    时间: 2005-4-14 06:31
标题: My consideration.

Let q(n,k) be the probability that not all safes can be opened, i.e., q(n,k)=1-p(n,k). Then in the total C(n,k) n! combinations, there are C(n,k) n!q(n,k) combinations that not all safes can be opened.  For  1<=i <=n-k, there are  C(n,i)i!(1-q(n-i, k))C(n-i, k) (n-i)! cases in which exactly i safes can not be opened (these i safes form some cycles...).
 
Therefore, we have:
 
    C(n,k) n!q(n,k)=C(n,1)1!(1-q(n-1, k))C(n-1, k)(n-1)!+....+C(n, n-k)(1-q(k,k))C(k,k)k!.
 
 
Now we are left to prove by induction on n-k that q(n,k)=1-k/n, which can be done easily.
www.ddhw.com

 

作者: fzy    时间: 2005-4-14 06:42
标题: 回复:回复:回复:保 险 箱 和 钥 匙

It is k/n. I did use induction (and still hate it!) There should be a more beautiful proof for such a beautiful result.www.ddhw.com

I tried 置换群 and it does not work (for me) either.www.ddhw.com

 

作者: 乱弹    时间: 2005-4-14 06:56
标题: Some typ0...

C(n,k) n!q(n,k)=C(n,1)1!(1-q(n-1, k))C(n-1, k)(n-1)!+....+C(n, n-k)(n-k)!(1-q(k,k))C(k,k)k!.

 

I did use 置换群, but not much.

www.ddhw.com

 


作者: fzy    时间: 2005-4-14 18:09
标题: A complete and simple solution

Still let P(n,k) be the probability of opening all n boxes given k keys. We first open one box and get another key. The probability of getting a new key (not already have) is (n-k)/n. So we have this simple recursive relation:
 
P(n,k) = k/n * P(n-1,k-1) + (n-k)/n * P(n-1,k)
 
From this and a double induction on n and k (DOUBLE! ) we get P(n,k) = k/n.
www.ddhw.com

 

作者: 乱弹    时间: 2005-4-14 20:02
标题: It is pretty pretty.[@};-][@};-]

To save yourself from double induction, you can try it on n+k. Anyway, it is not important.
www.ddhw.com

 

作者: 野 菜 花    时间: 2005-4-14 21:22
标题: Great job! [@};-][@};-]

  Great job!





作者: 野 菜 花    时间: 2005-4-14 21:27
标题: 好 厉 害 , 佩 服 ! [@};-][@};-]

I thought this question could last a little longer. It seems I could not find any problems which can beat you
www.ddhw.com

 





欢迎光临 珍珠湾ART (http://zzwav.com/) Powered by Discuz! X3