Big-O算法分析
我想说这不是家庭作业的问题。它只是 USACO 网站上学习动态规划概念的在线教程资源。 在资源中,给出了一个问题如下。
问题: 一个多达 10000 个整数的序列,( 0 < 整数 < 100,000),最大递减子序列是多少?
伙计们,给出了体面的递归方法
1 #include <stdio.h>
2 long n, sequence[10000];
3 main () {
4 FILE *in, *out;
5 int i;
6 in = fopen ("input.txt", "r");
7 out = fopen ("output.txt", "w");
8 fscanf(in, "%ld", &n);
9 for (i = 0; i < n; i++) fscanf(in, "%ld", &sequence[i]);
10 fprintf (out, "%d\n", check (0, 0, 99999));
11 exit (0);
12 }
13 check (start, nmatches, smallest) {
14 int better, i, best=nmatches;
15 for (i = start; i < n; i++) {
16 if (sequence[i] < smallest) {
17 better = check (i, nmatches+1, sequence[i]);
18 if (better > best) best = better;
19 }
20 }
21 return best;
22 }
,我不擅长算法分析。请您告诉我在最坏情况下尽可能严格的递归枚举解决方案的 Big-O 表示法是什么。我个人的想法是O(N^N),但是我没有信心。因为在N<=100下运行时间还是可以接受的。肯定有问题。请帮我。谢谢。
在USACO网站上,给出了O(n^2)的动态规划方法如下。
1 #include <stdio.h>
2 #define MAXN 10000
3 main () {
4 long num[MAXN], bestsofar[MAXN];
5 FILE *in, *out;
6 long n, i, j, longest = 0;
7 in = fopen ("input.txt", "r");
8 out = fopen ("output.txt", "w");
9 fscanf(in, "%ld", &n);
10 for (i = 0; i < n; i++) fscanf(in, "%ld", &num[i]);
11 bestsofar[n-1] = 1;
12 for (i = n-1-1; i >= 0; i--) {
13 bestsofar[i] = 1;
14 for (j = i+1; j < n; j++) {
15 if (num[j] < num[i] && bestsofar[j] >= bestsofar[i]) {
16 bestsofar[i] = bestsofar[j] + 1;
17 if (bestsofar[i] > longest) longest = bestsofar[i];
18 }
19 }
20 }
21 fprintf(out, "bestsofar is %d\n", longest);
22 exit(0);
23 }
I would say it's not a homework problem. It's just a tutorial resource online to learn the dynamic programming concepts from USACO website.
In the resource, a problem was given as follows.
Question:
A sequcen of as many as 10000 integers, ( 0 < integer < 100,000), what is the maximum decreasing subsequence?
The decent recursive approach was given
1 #include <stdio.h>
2 long n, sequence[10000];
3 main () {
4 FILE *in, *out;
5 int i;
6 in = fopen ("input.txt", "r");
7 out = fopen ("output.txt", "w");
8 fscanf(in, "%ld", &n);
9 for (i = 0; i < n; i++) fscanf(in, "%ld", &sequence[i]);
10 fprintf (out, "%d\n", check (0, 0, 99999));
11 exit (0);
12 }
13 check (start, nmatches, smallest) {
14 int better, i, best=nmatches;
15 for (i = start; i < n; i++) {
16 if (sequence[i] < smallest) {
17 better = check (i, nmatches+1, sequence[i]);
18 if (better > best) best = better;
19 }
20 }
21 return best;
22 }
Guys, I am not good at the algorithmic analysis. Would you please tell me what's the Big-O notation to this recursive enumeration solution in worst case as tight as possible. My personal thought would be O(N^N), but I have no confidence. Because the runtime is still acceptable under N <= 100. There must be something wrong. Please help me. Thank you.
In the USACO website, it gives the dynamic programming approach in O(n^2) as follows.
1 #include <stdio.h>
2 #define MAXN 10000
3 main () {
4 long num[MAXN], bestsofar[MAXN];
5 FILE *in, *out;
6 long n, i, j, longest = 0;
7 in = fopen ("input.txt", "r");
8 out = fopen ("output.txt", "w");
9 fscanf(in, "%ld", &n);
10 for (i = 0; i < n; i++) fscanf(in, "%ld", &num[i]);
11 bestsofar[n-1] = 1;
12 for (i = n-1-1; i >= 0; i--) {
13 bestsofar[i] = 1;
14 for (j = i+1; j < n; j++) {
15 if (num[j] < num[i] && bestsofar[j] >= bestsofar[i]) {
16 bestsofar[i] = bestsofar[j] + 1;
17 if (bestsofar[i] > longest) longest = bestsofar[i];
18 }
19 }
20 }
21 fprintf(out, "bestsofar is %d\n", longest);
22 exit(0);
23 }
如果你对这篇内容有疑问,欢迎到本站社区发帖提问 参与讨论,获取更多帮助,或者扫码二维码加入 Web 技术交流群。

绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱,如果其他用户或者作者回复了您的评论,将不能在第一时间通知您!
发布评论
评论(1)
只要看看你调用该函数时使用了什么样的参数即可。第一个决定第三个(顺便说一句,这意味着您需要第三个参数)。第一个范围介于 0 和 n 之间。第二个比第一个小。这意味着您最多对该函数有 n^2 次不同的调用。
现在的问题是您使用相同的参数调用该函数多少次。答案很简单:您实际上生成了每个递减子序列。这意味着对于序列 N、N-1、N-2...,您将生成 2^N 个序列。很差,对吧(如果你想尝试我给你的序列)?
但是,如果您使用您应该已经阅读过的记忆技术,您可以将复杂度提高到 N^3(每次调用该函数时最多执行 n 次操作,不同的调用是 N^2 并且记忆允许您只支付一次对于不同的呼叫)。
Just look at with what kind of parameters you call the function. The first determines the third (which btw means you needed have the third parameter). The first ranges between 0 and n. The second one is smaller than the first. This means that you have at most n^2 different calls to the function.
Now comes the question how many times you call the function with the same parameters. And the answer is simple: you actually generate every single decreasing subsequece. This means that for the sequence N, N-1, N-2, ... you will generate 2^N sequences. Pretty poor, right (if you want experiment with the sequence I have given you)?
However if you use the memoization technique you should have already read about, you can improve the complexity to N^3 (at most n operations in every call to the function, the different calls are N^2 and memoization allows you to pay only once for a different call).