Median

http://pat.zju.edu.cn/contests/pat-a-practise/1029

二分搜索median。

代码一： 时间复杂度log(m  + n)。

思路：比较两个数组的中间元素，然后可以排除掉其中一个数组的一半。

假设数组A, B, 我们要找的是A, B合并为一个数组后的第median个元素。令A, B中间元素分别为A[mid1], B[mid2]：

如果 A[mid1] == B[mid2]，则结果就为A[mid1];

如果 A[mid1] < B[mid2]，这时B[mid2]的前面一共有 mid1 + mid2 + 1个数字，

此时需要判断mid1 + mid2 + 1和median 的大小：

如果前者大，意味着我们要找的数字就在这 mid1 + mid2 + 1个数字中，从而我们可以排除B数组中的后一半；

如果后者大，意味着我们要找的数字不在这 mid1 + mid2 + 1个数字中，从而我们可以排除A 数组中的前一半。

当A[mid1] > B[mid2]时情况类似。

#include <iostream>
#include <cstdio>
#include <vector>
#include <cstring>
#include <algorithm>
#include <stack>
using namespace std;

long long *a, *b;

long long findMedian(long long *v1, long long *v2, int len1, int len2, int indexOfMid)
{
	if(len1 == 0) return v2[indexOfMid - 1];
	if(len2 == 0) return v1[indexOfMid - 1];
	int mid1 = (len1 - 1) >> 1, mid2 = (len2 - 1) >> 1;
	int index = mid1 + mid2 + 1;
	if(v1[mid1] == v2[mid2]) return v1[mid1];
	if(v1[mid1] < v2[mid2])
	{		
		if(index < indexOfMid) return findMedian(v1 + mid1 + 1, v2, len1 - mid1 - 1, len2, indexOfMid - mid1 - 1);
		else return findMedian(v1, v2, len1, mid2, indexOfMid);
	}
	else
	{
		if(index < indexOfMid) return findMedian(v1, v2 + mid2 + 1, len1, len2 - mid2 - 1, indexOfMid - mid2 - 1);
		return findMedian(v1, v2, mid1, len2, indexOfMid);
	}
}

int main()
{	
	int m, n, i;
	long long x;
	scanf("%d", &m);
	a = new long long[m + 10];
	for(i = 0; i < m; i ++)
	{
		scanf("%lld", &a[i]);
	}
	
	scanf("%d", &n);
	b = new long long[n + 10];
	for(i = 0; i < n; i ++)
	{
		scanf("%lld", &b[i]);
	}
	int mid = (m + n + 1) >> 1;
	long long ans = findMedian(a, b, m, n, mid);
	printf("%lld\n", ans);
	delete a;
	delete b;
	return 0;
}

代码二：时间复杂度logm*logn(想复杂了)

#include <iostream>
#include <cstdio>
#include <vector>
#include <cstring>
#include <algorithm>
#include <stack>
using namespace std;

long long *a, *b;

int binaryS1(long long *v, long long target, int size) // find first smaller than target
{
	int start = 0, end = size - 1;
	while(start < end)
	{
		int mid = (start + end) >> 1;
		if(v[mid] >= target) end = mid - 1;
		else start = mid + 1;
	}
	if(v[start] >= target) start -- ;
	return start;
}

int binaryS2(long long *v, long long target, int size) // find first larger than target
{
	int start = 0, end = size - 1;
	while(start < end)
	{
		int mid = (start + end) >> 1;
		if(v[mid] <= target) start = mid + 1;
		else end = mid - 1;
	}
	if(v[start] <= target) start ++ ;
	return start;
}

bool isFind = false;

long long findMedian(long long *v1, long long *v2, int size1, int size2)
{
	int start = 0, end = size1 - 1;
	int median = (size1 + size2) >> 1;
	if((size1 + size2) % 2 == 1) median ++ ;
	while(start <= end)
	{
		int mid = (start + end) >> 1;
		int index1 = binaryS1(v2, v1[mid], size2);
		int index2 = binaryS2(v2, v1[mid], size2);
		int posMin = mid + 1 + index1 + 1;
		int posMax = mid + 1 + index2;
		if(posMin <= median && posMax >= median) 
		{
			isFind = true;
			return v1[mid];
		}
		else if(posMin > median) end = mid - 1;
		else start = mid + 1;
	}
	return -1;
}


int main()
{	
	int m, n, i;
	long long x;
	scanf("%d", &m);
	a = new long long[m + 10];
	for(i = 0; i < m; i ++)
	{
		scanf("%lld", &a[i]);
	}
	
	scanf("%d", &n);
	b = new long long[n + 10];
	for(i = 0; i < n; i ++)
	{
		scanf("%lld", &b[i]);
	}
	long long ans = findMedian(a, b, m, n);
	if(isFind == false) ans = findMedian(b, a, n, m);
	printf("%lld\n", ans);
	delete a;
	delete b;
	return 0;
}

当然这题还有O(m + n)的算法，即归并排序思想。