[HNU 10027] Longest Ordered Subsequence Extention

Longest Ordered Subsequence Extention
Time Limit: 1000ms, Special Time Limit:2500ms, Memory Limit:65536KB
Total submit users: 1521, Accepted users: 991
Problem 10027 : No special judgement
Problem description
A numeric sequence of ai is ordered if a1 < a2 < … < aN. Let the subsequence of the given numeric sequence (a1, a2, …, aN) be any sequence (ai1, ai2, …, aiK), where 1 <= i1 < i2 < … < iK <= N. For example, sequence (1, 7, 3, 5, 9, 4, 8) has ordered subsequences, e. g., (1, 7), (3, 4, 8) and many others. All longest ordered subsequences are of length 4, e. g., (1, 3, 5, 8).

Your program, when given the numeric sequence, must find the length of its longest ordered subsequence.

Please notice this problem is different with 10001,^_.

Input
The first line of input file contains the length of sequence N. The second line contains the elements of sequence - N integers in the range from 0 to 100000 each, separated by spaces. 1 <= N <= 50000

Output
Output file must contain a single integer - the length of the longest ordered subsequence of the given sequence.

Sample Input
7
1 7 3 5 9 4 8
Sample Output
4
Problem Source
HNU Contest

#include <stdio.h>
#include <algorithm>

const int MAX_N = 50005;

int a[MAX_N], dp[MAX_N], n;

template <typename T>
inline T read_int(T& n)
{
	bool f = 1; char ch = getchar();
	while (ch < '0' || ch > '9') { if (ch == '-') !f; ch = getchar(); }
	while (ch >= '0' && ch <= '9') { n = (n << 1) + (n << 3) + ch - '0'; ch = getchar(); }
	n = f ? n : ~(n-1);
}

template <typename T>
int m_lower_bound(T a[], int left, int right, T target)
{
	int l = left, r = right;
	while (l <= r)
	{
		int m = (l + r) >> 1;
		if (a[m] == target)
		{
			if (m == left || m != left && a[m-1] != target) return m;
			r = m - 1;	
		}
		else if (a[m] > target) r = m - 1; 
		else if (a[m] < target) l = m + 1;		
	}
	return l <= right ? l : -1;
}


int solve()
{
	int max_pos = 0;
	dp[max_pos] = a[0];
	for (int i = 1 ; i < n ; ++i) 
	{
		if (a[i] > dp[max_pos]) dp[++max_pos] = a[i];
		//else dp[m_lower_bound(dp, 0, max_pos, a[i])] = a[i];   // 比 STL 慢 15ms 
		else dp[std::lower_bound(dp, dp+max_pos+1, a[i])-dp] = a[i];	
	}
	return max_pos + 1;
}


int main()
{
	read_int(n);
	for (int i = 0 ; i < n ; ++i) read_int(a[i]);  
	printf("%d", solve());
	return 0;
}

算法思路：
dp + 二分查找，时间复杂度 nlogn。
如果这里不能二分，则还是 O(n^2) 对于 5w 的数据，得到的是 5w * 5w = 2.5 * 10 ^ 9, 10 ** 6 大约是 0.1s，2.5 * 10 ^ 9 需要 250s。

思路：
状态设置为 dp[i] 表示第长度为 i - 1 的最长上升子序列的最后一位的最小值（这里从数组下标为0处开始使用，所以是长度为 i - 1 ）

另外：
开快读比 scanf 快 200ms 。
STL 的 lower_bound 比自己写的快 15ms。