LC 382. 链表随机节点

题目描述

这是 LeetCode 上的 382. 链表随机节点 ,难度为 中等

Tag :「链表」、「模拟」、「蓄水池抽样」

给你一个单链表,随机选择链表的一个节点,并返回相应的节点值。每个节点 被选中的概率一样

实现 Solution 类:

  • Solution(ListNode head) 使用整数数组初始化对象。
  • int getRandom() 从链表中随机选择一个节点并返回该节点的值。链表中所有节点被选中的概率相等。

示例:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
输入
["Solution", "getRandom", "getRandom", "getRandom", "getRandom", "getRandom"]
[[[1, 2, 3]], [], [], [], [], []]

输出
[null, 1, 3, 2, 2, 3]

解释
Solution solution = new Solution([1, 2, 3]);
solution.getRandom(); // 返回 1
solution.getRandom(); // 返回 3
solution.getRandom(); // 返回 2
solution.getRandom(); // 返回 2
solution.getRandom(); // 返回 3
// getRandom() 方法应随机返回 123中的一个,每个元素被返回的概率相等。

提示:

  • $链表中的节点数在范围 [1, 10^4] 内$
  • $-10^4 <= Node.val <= 10^4$
  • 至多调用 getRandom 方法 $10^4$ 次

进阶:

  • 如果链表非常大且长度未知,该怎么处理?
  • 你能否在不使用额外空间的情况下解决此问题?

模拟

由于链表长度只有 $10^4$,因此可以在初始化时遍历整条链表,将所有的链表值预处理到一个数组内。

在查询时随机一个下标,并将数组中对应下标内容返回出去。

代码(感谢 @Benhao 同学提供的其他语言版本):

1
2
3
4
5
6
7
8
9
10
11
12
13
14
class Solution {
List<Integer> list = new ArrayList<>();
Random random = new Random(20220116);
public Solution(ListNode head) {
while (head != null) {
list.add(head.val);
head = head.next;
}
}
public int getRandom() {
int idx = random.nextInt(list.size());
return list.get(idx);
}
}

-
1
2
3
4
5
6
7
8
9
10
class Solution:

def __init__(self, head: Optional[ListNode]):
self.nodes = []
while head:
self.nodes.append(head)
head = head.next

def getRandom(self) -> int:
return self.nodes[randint(0, len(self.nodes) - 1)].val

-
1
2
3
4
5
6
7
8
9
10
11
12
13
14
type Solution struct {
Nodes []int
}
func Constructor(head *ListNode) Solution {
nodes := make([]int, 0)
for head != nil {
nodes = append(nodes, head.Val)
head = head.Next
}
return Solution{nodes}
}
func (this *Solution) GetRandom() int {
return this.Nodes[rand.Intn(len(this.Nodes))]
}

  • 时间复杂度:令 $n$ 为链表长度,预处理数组的复杂度为 $O(n)$;随机获取某个值的复杂度为 $O(1)$
  • 空间复杂度:$O(n)$

蓄水池抽样

整理题意:总的样本数量未知,从所有样本中抽取若干个,要求每个样本被抽到的概率相等。

具体做法为:从前往后处理每个样本,每个样本成为答案的概率为 $\frac{1}{i}$,其中 $i$ 为样本编号(编号从 $1$ 开始),最终可以确保每个样本成为答案的概率均为 $\frac{1}{n}$(其中 $n$ 为样本总数)。

容易证明该做法的正确性,假设最终成为答案的样本编号为 $k$,那么 $k$ 成为答案的充要条件为「在遍历到 $k$ 时被选中」并且「遍历大于 $k$ 的所有元素时,均没有被选择(没有覆盖 $k$)」。

对应事件概率为:

首项 $\frac{1}{k}$ 为选中 $k$ 的概率,后面每项分别为编号为 $[k + 1, n]$ 的样本 不被选中 的概率。

化简得:

进一步抵消化简后,可得:

因此,在每一次 getRandom 时,从前往后处理每个节点,同时记录当前节点的编号,当处理到节点 $k$ 时,在 $[0, k)$ 范围内进行随机,若随机到结果为 $0$(发生概率为 $\frac{1}{k}$),则将节点 $k$ 的值存入答案,最后一次覆盖答案的节点即为本次抽样结果。

代码(感谢 @Benhao 同学提供的其他语言版本):

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
class Solution {
ListNode head;
Random random = new Random(20220116);
public Solution(ListNode _head) {
head = _head;
}
public int getRandom() {
int ans = 0, idx = 0;
ListNode t = head;
while (t != null && ++idx >= 0) {
if (random.nextInt(idx) == 0) ans = t.val;
t = t.next;
}
return ans;
}
}

-
1
2
3
4
5
6
7
8
9
10
11
12
class Solution:

def __init__(self, head: Optional[ListNode]):
self.root = head

def getRandom(self) -> int:
node, ans, i = self.root, None, 0
while node:
if not randint(0, i):
ans = node.val
node, i = node.next, i + 1
return ans

-
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
type Solution struct {
Root *ListNode
}
func Constructor(head *ListNode) Solution {
return Solution{head}
}
func (this *Solution) GetRandom() (ans int) {
for node, idx := this.Root, 1;node != nil; idx++ {
if rand.Intn(idx) == 0{
ans = node.Val
}
node = node.Next
}
return
}

  • 时间复杂度:令 $n$ 为链表长度,随机获取某个值的复杂度为 $O(n)$
  • 空间复杂度:$O(1)$

最后

这是我们「刷穿 LeetCode」系列文章的第 No.382 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。

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。

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

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。


本博客所有文章除特别声明外,均采用 CC BY-SA 4.0 协议 ,转载请注明出处!