Skip to content

Commit dd48c2b

Browse files
Merge pull request SharingSource#309 from SharingSource/ac_oier
✨feat: Add 373
2 parents ff512be + ceeb4f0 commit dd48c2b

File tree

4 files changed

+265
-0
lines changed

4 files changed

+265
-0
lines changed

Index/二分.md

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -20,6 +20,7 @@
2020
| [354. 俄罗斯套娃信封问题](https://leetcode-cn.com/problems/russian-doll-envelopes/) | [LeetCode 题解链接](https://leetcode-cn.com/problems/russian-doll-envelopes/solution/zui-chang-shang-sheng-zi-xu-lie-bian-xin-6s8d/) | 困难 | 🤩🤩🤩 |
2121
| [363. 矩形区域不超过 K 的最大数值和](https://leetcode-cn.com/problems/max-sum-of-rectangle-no-larger-than-k/) | [LeetCode 题解链接](https://leetcode-cn.com/problems/max-sum-of-rectangle-no-larger-than-k/solution/gong-shui-san-xie-you-hua-mei-ju-de-ji-b-dh8s/) | 困难 | 🤩🤩🤩 |
2222
| [367. 有效的完全平方数](https://leetcode-cn.com/problems/valid-perfect-square/) | [LeetCode 题解链接](https://leetcode-cn.com/problems/valid-perfect-square/solution/gong-shui-san-xie-yi-ti-shuang-jie-er-fe-g5el/) | 简单 | 🤩🤩🤩🤩🤩 |
23+
| [373. 查找和最小的K对数字](https://leetcode-cn.com/problems/find-k-pairs-with-smallest-sums/) | [LeetCode 题解链接](https://leetcode-cn.com/problems/find-k-pairs-with-smallest-sums/solution/gong-shui-san-xie-duo-lu-gui-bing-yun-yo-pgw5/) | 中等 | 🤩🤩🤩🤩🤩 |
2324
| [374. 猜数字大小](https://leetcode-cn.com/problems/guess-number-higher-or-lower/) | [LeetCode 题解链接](https://leetcode-cn.com/problems/guess-number-higher-or-lower/solution/gong-shui-san-xie-shi-yong-jiao-hu-han-s-tocm/) | 简单 | 🤩🤩🤩 |
2425
| [441. 排列硬币](https://leetcode-cn.com/problems/arranging-coins/) | [LeetCode 题解链接](https://leetcode-cn.com/problems/arranging-coins/solution/gong-shui-san-xie-yi-ti-shuang-jie-shu-x-sv9o/) | 简单 | 🤩🤩🤩 |
2526
| [475. 供暖器](https://leetcode-cn.com/problems/heaters/) | [LeetCode 题解链接](https://leetcode-cn.com/problems/heaters/solution/gong-shui-san-xie-er-fen-shuang-zhi-zhen-mys4/) | 中等 | 🤩🤩🤩🤩 |

Index/堆.md

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -5,6 +5,7 @@
55
| [264. 丑数 II](https://leetcode-cn.com/problems/ugly-number-ii/) | [LeetCode 题解链接](https://leetcode-cn.com/problems/ugly-number-ii/solution/gong-shui-san-xie-yi-ti-shuang-jie-you-x-3nvs/) | 中等 | 🤩🤩🤩 |
66
| [295. 数据流的中位数](https://leetcode-cn.com/problems/find-median-from-data-stream/) | [LeetCode 题解链接](https://leetcode-cn.com/problems/find-median-from-data-stream/solution/gong-shui-san-xie-jing-dian-shu-ju-jie-g-pqy8/) | 中等 | 🤩🤩🤩🤩 |
77
| [313. 超级丑数](https://leetcode-cn.com/problems/super-ugly-number/) | [LeetCode 题解链接](https://leetcode-cn.com/problems/super-ugly-number/solution/gong-shui-san-xie-yi-ti-shuang-jie-you-x-jyow/) | 中等 | 🤩🤩🤩 |
8+
| [373. 查找和最小的K对数字](https://leetcode-cn.com/problems/find-k-pairs-with-smallest-sums/) | [LeetCode 题解链接](https://leetcode-cn.com/problems/find-k-pairs-with-smallest-sums/solution/gong-shui-san-xie-duo-lu-gui-bing-yun-yo-pgw5/) | 中等 | 🤩🤩🤩🤩🤩 |
89
| [407. 接雨水 II](https://leetcode-cn.com/problems/trapping-rain-water-ii/) | [LeetCode 题解链接](https://leetcode-cn.com/problems/trapping-rain-water-ii/solution/gong-shui-san-xie-jing-dian-dijkstra-yun-13ik/) | 困难 | 🤩🤩🤩🤩 |
910
| [451. 根据字符出现频率排序](https://leetcode-cn.com/problems/sort-characters-by-frequency/) | [LeetCode 题解链接](https://leetcode-cn.com/problems/sort-characters-by-frequency/solution/gong-shui-san-xie-shu-ju-jie-gou-yun-yon-gst9/) | 中等 | 🤩🤩🤩🤩 |
1011
| [480. 滑动窗口中位数](https://leetcode-cn.com/problems/sliding-window-median/) | [LeetCode 题解链接](https://leetcode-cn.com/problems/sliding-window-median/solution/xiang-jie-po-su-jie-fa-you-xian-dui-lie-mo397/) | 困难 | 🤩🤩🤩🤩 |

Index/多路归并.md

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -3,5 +3,6 @@
33
| [21. 合并两个有序链表](https://leetcode-cn.com/problems/merge-two-sorted-lists/) | [LeetCode 题解链接](https://leetcode-cn.com/problems/merge-two-sorted-lists/solution/shua-chuan-lc-shuang-zhi-zhen-jie-fa-sha-b22z/) | 简单 | 🤩🤩🤩🤩🤩 |
44
| [264. 丑数 II](https://leetcode-cn.com/problems/ugly-number-ii/) | [LeetCode 题解链接](https://leetcode-cn.com/problems/ugly-number-ii/solution/gong-shui-san-xie-yi-ti-shuang-jie-you-x-3nvs/) | 中等 | 🤩🤩🤩🤩 |
55
| [313. 超级丑数](https://leetcode-cn.com/problems/super-ugly-number/) | [LeetCode 题解链接](https://leetcode-cn.com/problems/super-ugly-number/solution/gong-shui-san-xie-yi-ti-shuang-jie-you-x-jyow/) | 中等 | 🤩🤩🤩 |
6+
| [373. 查找和最小的K对数字](https://leetcode-cn.com/problems/find-k-pairs-with-smallest-sums/) | [LeetCode 题解链接](https://leetcode-cn.com/problems/find-k-pairs-with-smallest-sums/solution/gong-shui-san-xie-duo-lu-gui-bing-yun-yo-pgw5/) | 中等 | 🤩🤩🤩🤩🤩 |
67
| [786. 第 K 个最小的素数分数](https://leetcode-cn.com/problems/k-th-smallest-prime-fraction/) | [LeetCode 题解链接](https://leetcode-cn.com/problems/k-th-smallest-prime-fraction/solution/gong-shui-san-xie-yi-ti-shuang-jie-you-x-8ymk/) | 中等 | 🤩🤩🤩 |
78

Lines changed: 262 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,262 @@
1+
### 题目描述
2+
3+
这是 LeetCode 上的 **[373. 查找和最小的K对数字](https://leetcode-cn.com/problems/find-k-pairs-with-smallest-sums/solution/gong-shui-san-xie-duo-lu-gui-bing-yun-yo-pgw5/)** ,难度为 **中等**
4+
5+
Tag : 「优先队列」、「二分」、「多路归并」
6+
7+
8+
9+
给定两个以升序排列的整数数组 `nums1``nums2` , 以及一个整数 `k` 。
10+
11+
定义一对值 $(u,v)$,其中第一个元素来自 `nums1`,第二个元素来自 `nums2`
12+
13+
请找到和最小的 `k` 个数对 $(u_1,v_1),  (u_2,v_2)  ...  (u_k,v_k)$ 。 
14+
15+
示例 1:
16+
```
17+
输入: nums1 = [1,7,11], nums2 = [2,4,6], k = 3
18+
19+
输出: [1,2],[1,4],[1,6]
20+
21+
解释: 返回序列中的前 3 对数:
22+
[1,2],[1,4],[1,6],[7,2],[7,4],[11,2],[7,6],[11,4],[11,6]
23+
```
24+
示例 2:
25+
```
26+
输入: nums1 = [1,1,2], nums2 = [1,2,3], k = 2
27+
28+
输出: [1,1],[1,1]
29+
30+
解释: 返回序列中的前 2 对数:
31+
  [1,1],[1,1],[1,2],[2,1],[1,2],[2,2],[1,3],[1,3],[2,3]
32+
```
33+
示例 3:
34+
```
35+
输入: nums1 = [1,2], nums2 = [3], k = 3
36+
37+
输出: [1,3],[2,3]
38+
39+
解释: 也可能序列中所有的数对都被返回:[1,3],[2,3]
40+
```
41+
42+
提示:
43+
* $1 <= nums1.length, nums2.length <= 10^4$
44+
* $-10^9 <= nums1[i], nums2[i] <= 10^9$
45+
* $nums1, nums2 均为升序排列$
46+
* $1 <= k <= 1000$
47+
48+
---
49+
50+
### 基本分析
51+
52+
这道题和 [(题解) 786. 第 K 个最小的素数分数](https://leetcode-cn.com/problems/k-th-smallest-prime-fraction/solution/gong-shui-san-xie-yi-ti-shuang-jie-you-x-8ymk/) 几乎是一模一样,先做哪一道都是一样的,难度上没有区别 🤣
53+
54+
最常规的做法是使用「多路归并」,还不熟悉「多路归并」的同学,建议先学习前置🧀:[多路归并入门](https://mp.weixin.qq.com/s?__biz=MzU4NDE3MTEyMA==&mid=2247490029&idx=1&sn=bba9ddff88d247db310406ee418d5a15&chksm=fd9cb2f2caeb3be4b1f84962677337dcb5884374e5b6b80340834eaff79298d11151da2dd5f7&token=252055586&lang=zh_CN#rd),里面讲述了如何从「朴素优先队列」往「多路归并」进行转换。
55+
56+
---
57+
58+
### 多路归并
59+
60+
令 $nums1$ 的长度为 $n$,$nums2$ 的长度为 $m$,所有的点对数量为 $n * m$。
61+
62+
其中每个 $nums1[i]$ 参与所组成的点序列为:
63+
64+
$$
65+
[(nums1[0], nums2[0]), (nums1[0], nums2[1]), ..., (nums1[0], nums2[m - 1])]\\
66+
[(nums1[1], nums2[0]), (nums1[1], nums2[1]), ..., (nums1[1], nums2[m - 1])]\\
67+
...\\
68+
[(nums1[n - 1], nums2[0]), (nums1[n - 1], nums2[1]), ..., (nums1[n - 1], nums2[m - 1])]\\
69+
$$
70+
71+
由于 $nums1$ 和 $nums2$ 均已按升序排序,因此每个 $nums1[i]$ 参与构成的点序列也为升序排序,这引导我们使用「多路归并」来进行求解。
72+
73+
具体的,起始我们将这 $n$ 个序列的首位元素(点对)以二元组 $(i, j)$ 放入优先队列(小根堆),其中 $i$ 为该点对中 $nums1[i]$ 的下标,$j$ 为该点对中 $nums2[j]$ 的下标,这步操作的复杂度为 $O(n\log{n})$。这里也可以得出一个小优化是:我们始终确保 $nums1$ 为两数组中长度较少的那个,然后通过标识位来记录是否发生过交换,确保答案的点顺序的正确性。
74+
75+
每次从优先队列(堆)中取出堆顶元素(含义为当前未被加入到答案的所有点对中的最小值),加入答案,并将该点对所在序列的下一位(如果有)加入优先队列中。
76+
77+
举个 🌰,首次取出的二元组为 $(0, 0)$,即点对 $(nums1[0], nums2[0])$,取完后将序列的下一位点对 $(nums1[0], nums2[1])$ 以二元组 $(0, 1)$ 形式放入优先队列。
78+
79+
可通过「反证法」证明,每次这样的「取当前,放入下一位」的操作,可以确保当前未被加入答案的所有点对的最小值必然在优先队列(堆)中,即前 $k$ 个出堆的元素必然是所有点对的前 $k$ 小的值。
80+
81+
**代码(感谢 [@Benhao](/u/himymben/) 同学提供的其他语言版本):**
82+
```Java
83+
class Solution {
84+
boolean flag = true;
85+
public List<List<Integer>> kSmallestPairs(int[] nums1, int[] nums2, int k) {
86+
List<List<Integer>> ans = new ArrayList<>();
87+
int n = nums1.length, m = nums2.length;
88+
if (n > m && !(flag = false)) return kSmallestPairs(nums2, nums1, k);
89+
PriorityQueue<int[]> q = new PriorityQueue<>((a,b)->(nums1[a[0]]+nums2[a[1]])-(nums1[b[0]]+nums2[b[1]]));
90+
for (int i = 0; i < Math.min(n, k); i++) q.add(new int[]{i, 0});
91+
while (ans.size() < k && !q.isEmpty()) {
92+
int[] poll = q.poll();
93+
int a = poll[0], b = poll[1];
94+
ans.add(new ArrayList<>(){{
95+
add(flag ? nums1[a] : nums2[b]);
96+
add(flag ? nums2[b] : nums1[a]);
97+
}});
98+
if (b + 1 < m) q.add(new int[]{a, b + 1});
99+
}
100+
return ans;
101+
}
102+
}
103+
```
104+
-
105+
```Python3
106+
class Solution:
107+
def kSmallestPairs(self, nums1: List[int], nums2: List[int], k: int) -> List[List[int]]:
108+
flag, ans = (n := len(nums1)) > (m := len(nums2)), []
109+
if flag:
110+
n, m, nums1, nums2 = m, n, nums2, nums1
111+
pq = []
112+
for i in range(min(n, k)):
113+
heapq.heappush(pq, (nums1[i] + nums2[0], i, 0))
114+
while len(ans) < k and pq:
115+
_, a, b = heapq.heappop(pq)
116+
ans.append([nums2[b], nums1[a]] if flag else [nums1[a], nums2[b]])
117+
if b + 1 < m:
118+
heapq.heappush(pq, (nums1[a] + nums2[b + 1], a, b + 1))
119+
return ans
120+
```
121+
-
122+
```Golang
123+
func kSmallestPairs(nums1 []int, nums2 []int, k int) [][]int {
124+
n, m, ans := len(nums1), len(nums2), [][]int{}
125+
flag := n > m
126+
if flag {
127+
n, m, nums1, nums2 = m, n, nums2, nums1
128+
}
129+
if n > k {
130+
n = k
131+
}
132+
pq := make(hp, n)
133+
for i := 0; i < n; i++ {
134+
pq[i] = []int{nums1[i] + nums2[0], i, 0}
135+
}
136+
heap.Init(&pq)
137+
for pq.Len() > 0 && len(ans) < k {
138+
poll := heap.Pop(&pq).([]int)
139+
a, b := poll[1], poll[2]
140+
if flag{
141+
ans = append(ans, []int{nums2[b], nums1[a]})
142+
}else{
143+
ans = append(ans, []int{nums1[a], nums2[b]})
144+
}
145+
if b < m - 1 {
146+
heap.Push(&pq, []int{nums1[a] + nums2[b + 1], a, b + 1})
147+
}
148+
}
149+
return ans
150+
}
151+
// 最小堆模板
152+
type hp [][]int
153+
func (h hp) Len() int { return len(h) }
154+
func (h hp) Less(i, j int) bool { return h[i][0] < h[j][0] }
155+
func (h hp) Swap(i, j int) { h[i], h[j] = h[j], h[i] }
156+
func (h *hp) Push(v interface{}) { *h = append(*h, v.([]int)) }
157+
func (h *hp) Pop() interface{} { a := *h; v := a[len(a)-1]; *h = a[:len(a)-1]; return v }
158+
```
159+
* 时间复杂度:令 $M$ 为 $n$、$m$ 和 $k$ 三者中的最小值,复杂度为 $O(M + k) * \log{M})$
160+
* 空间复杂度:$O(M)$
161+
162+
---
163+
164+
### 二分
165+
166+
我们还能够使用多次「二分」来做。
167+
168+
假设我们将所有「数对和」按照升序排序,两端的值分别为 $l = nums1[0] + nums2[0]$ 和 $r = nums1[n - 1] + nums2[m - 1]$。
169+
170+
因此我们可以在值域 $[l, r]$ 上进行二分,找到第一个满足「点对和小于等于 $x$ 的,且数量超过 $k$ 的值 $x$」。
171+
172+
之所以能够二分,是因为 $x$ 所在的点对和数轴上具有二段性:
173+
174+
* 点对和小于 $x$ 的点对数量少于 $k$ 个;
175+
* 点对和大于等于 $x$ 的点对数量大于等于 $k$ 个。
176+
177+
判定小于等于 $x$ 的点对数量是否大于等于 $k$ 个这一步可直接使用循环来做,由于二分是从中间值开始,这一步不会出现跑满两层循环的情况。
178+
179+
当二分出第 $k$ 小的值为 $x$ 后,由于存在不同点对的点对和值相等,我们需要先将所有点对和小于等于 $x$ 的值加入答案,然后酌情把值等于 $x$ 的点对加入答案,知道满足答案数量为 $k$。
180+
181+
找值为 $x$ 的所有点对这一步,可以通过枚举 $nums1[i]$,然后在 $nums2$ 上二分目标值 $x - nums1[i]$ 的左右端点来做。
182+
183+
最后,在所有处理过程中,我们都可以利用答案数组的大小与 $k$ 的关系做剪枝。
184+
185+
代码:
186+
```Java
187+
class Solution {
188+
int[] nums1, nums2;
189+
int n, m;
190+
public List<List<Integer>> kSmallestPairs(int[] n1, int[] n2, int k) {
191+
nums1 = n1; nums2 = n2;
192+
n = nums1.length; m = nums2.length;
193+
List<List<Integer>> ans = new ArrayList<>();
194+
int l = nums1[0] + nums2[0], r = nums1[n - 1] + nums2[m - 1];
195+
while (l < r) {
196+
int mid = l + r >> 1;
197+
if (check(mid, k)) r = mid;
198+
else l = mid + 1;
199+
}
200+
int x = r;
201+
for (int i = 0; i < n; i++) {
202+
for (int j = 0; j < m; j++) {
203+
if (nums1[i] + nums2[j] < x) {
204+
List<Integer> temp = new ArrayList<>();
205+
temp.add(nums1[i]); temp.add(nums2[j]);
206+
ans.add(temp);
207+
} else break;
208+
}
209+
}
210+
for (int i = 0; i < n && ans.size() < k; i++) {
211+
int a = nums1[i], b = x - a;
212+
int c = -1, d = -1;
213+
l = 0; r = m - 1;
214+
while (l < r) {
215+
int mid = l + r >> 1;
216+
if (nums2[mid] >= b) r = mid;
217+
else l = mid + 1;
218+
}
219+
if (nums2[r] != b) continue;
220+
c = r;
221+
l = 0; r = m - 1;
222+
while (l < r) {
223+
int mid = l + r + 1 >> 1;
224+
if (nums2[mid] <= b) l = mid;
225+
else r = mid - 1;
226+
}
227+
d = r;
228+
for (int p = c; p <= d && ans.size() < k; p++) {
229+
List<Integer> temp = new ArrayList<>();
230+
temp.add(a); temp.add(b);
231+
ans.add(temp);
232+
}
233+
}
234+
return ans;
235+
}
236+
boolean check(int x, int k) {
237+
int ans = 0;
238+
for (int i = 0; i < n && ans < k; i++) {
239+
for (int j = 0; j < m && ans < k; j++) {
240+
if (nums1[i] + nums2[j] <= x) ans++;
241+
else break;
242+
}
243+
}
244+
return ans >= k;
245+
}
246+
}
247+
```
248+
* 时间复杂度:假设点对和的值域大小范围为 $M$,第一次二分的复杂度为 $O((n * m) * \log{M})$;统计点对和值小于目标值 $x$ 的复杂度为 $O(n * m)$;统计所有点对和等于目标值的复杂度为 $O(\max(n * \log{m}, k))$(整个处理过程中利用了大小关系做了剪枝,大多循环都不会跑满,实际计算量会比理论分析的要低)
249+
* 空间复杂度:$O(k)$
250+
251+
---
252+
253+
### 最后
254+
255+
这是我们「刷穿 LeetCode」系列文章的第 `No.373` 篇,系列开始于 2021/01/01,截止于起始日 LeetCode 上共有 1916 道题目,部分是有锁题,我们将先把所有不带锁的题目刷完。
256+
257+
在这个系列文章里面,除了讲解解题思路以外,还会尽可能给出最为简洁的代码。如果涉及通解还会相应的代码模板。
258+
259+
为了方便各位同学能够电脑上进行调试和提交代码,我建立了相关的仓库:https://github.com/SharingSource/LogicStack-LeetCode
260+
261+
在仓库地址里,你可以看到系列文章的题解链接、系列文章的相应代码、LeetCode 原题链接和其他优选题解。
262+

0 commit comments

Comments
 (0)