class Solution:
def reverseKGroup(self, head: Optional[ListNode], k: int) -> Optional[ListNode]:
# 翻转链表前k项
def reverse(head, k):
pre, cur = None, head
while cur and k:
temp = cur.next
cur.next = pre
pre = cur
cur = temp
k -= 1
return pre, head # pre为头,head为尾
dummy = ListNode(0, head)
pre = cur = dummy
count = k # 用于重复计数
while cur.next and count:
count -= 1
cur = cur.next
if count == 0:
temp = cur.next # 存一下段后节点
pre.next, cur = reverse(pre.next, k) # 连接段前+翻转
cur.next = temp # 连上段后节点
pre = cur # 更新pre指针
count = k # 恢复count继续遍历
return dummy.next
138. 随机链表的复制 - 力扣(LeetCode)
路飞的题解真是太强啦!!优雅清晰简洁
哈希表
"""
# Definition for a Node.
class Node:
def __init__(self, x: int, next: 'Node' = None, random: 'Node' = None):
self.val = int(x)
self.next = next
self.random = random
"""
class Solution:
def copyRandomList(self, head: 'Optional[Node]') -> 'Optional[Node]':
if not head: return None
dic = {}
# 1. 复制各节点,并建立 “原节点 -> 新节点” 的 Map 映射
cur = head
while cur:
dic[cur] = Node(cur.val)
cur = cur.next
# 2. 构建新节点的 next 和 random 指向
cur = head
while cur:
dic[cur].next = dic.get(cur.next)
dic[cur].random = dic.get(cur.random)
cur = cur.next
# 3. 返回新链表的头节点
return dic[head]
拼接 + 拆分
class Solution:
def copyRandomList(self, head: 'Optional[Node]') -> 'Optional[Node]':
if not head: return None
dic = {}
# 1. 复制各节点,并构建拼接链表
cur = head
while cur:
temp = Node(cur.val)
temp.next = cur.next
cur.next = temp
cur = temp.next
# 2. 构建各新节点的 random 指向
cur = head
while cur:
if cur.random:
cur.next.random = cur.random.next
cur = cur.next.next
# 3. 拆分两链表
cur = newhead = head.next
pre = head
while cur.next:
pre.next = pre.next.next
cur.next = cur.next.next
pre = pre.next
cur = cur.next
pre.next = None # 单独处理原链表尾节点
return newhead # 返回新链表头节点
148. 排序链表 - 力扣(LeetCode)
归并排序(顶到底递归)
参考路飞题解
class Solution:
def sortList(self, head: Optional[ListNode]) -> Optional[ListNode]:
if not head or not head.next: return head
# 快慢指针分割链表
slow, fast = head, head.next
while fast and fast.next:
fast, slow = fast.next.next, slow.next
mid = slow.next # 右半部分的头节点
slow.next = None # 断开两部分
# 递归进行归并排序
left = self.sortList(head)
right = self.sortList(mid)
# 合并左右两个链表
dummy = cur = ListNode(0)
while left and right: # 根据大小依次插入新链表
if left.val < right.val:
cur.next = left
left = left.next
else:
cur.next = right
right = right.next
cur = cur.next
cur.next = left if left else right # 接上剩下的
return dummy.next
归并排序(底到顶合并)
class Solution:
def sortList(self, head: ListNode) -> ListNode:
# 合并两个有序链表
def merge(head1, head2):
dummy = cur = ListNode(0)
while head1 and head2:
if head1.val < head2.val:
cur.next = head1
head1 = head1.next
else:
cur.next = head2
head2 = head2.next
cur = cur.next
cur.next = head1 if head1 else head2
return dummy.next
# 如果只有一个节点直接返回head
if not head: return head
# 统计链表长度
lenth = 0
cur = head
while cur:
cur = cur.next
lenth += 1
# 开始循环合并
dummy = ListNode(0, head)
sublenth = 1
while sublenth < lenth:
pre, cur = dummy, dummy.next
while cur:
head1 = cur
for i in range(1, sublenth):
if cur.next:
cur = cur.next
else:
break # 如果还没找到head2说明不用合并,下一轮
head2 = cur.next
if not head2: break # 空就不合并了
cur.next = None # 断开第一段后
cur = head2
for i in range(1, sublenth):
if cur.next:
cur = cur.next
else:
break
temp = cur.next
cur.next = None # 断开第二段后
cur = temp
merged = merge(head1, head2) # 合并
pre.next = merged
while pre.next:
pre = pre.next # pre更新到合并后链表的最后
pre.next = temp # 重新连接第二段后
# 下一轮合并
sublenth *= 2
return dummy.next
23. 合并 K 个升序链表 - 力扣(LeetCode)
依次合并
class Solution:
def mergeKLists(self, lists: List[Optional[ListNode]]) -> Optional[ListNode]:
# 合并两个有序链表
def merge(head1, head2):
dummy = cur = ListNode(0)
while head1 and head2:
if head1.val < head2.val:
cur.next = head1
head1 = head1.next
else:
cur.next = head2
head2 = head2.next
cur = cur.next
cur.next = head1 if head1 else head2
return dummy.next
lenth = len(lists)
if lenth == 0:
return None
# 每遍历一个链表就合并掉
dummyhead = ListNode(0)
for i in range(0, len(lists)):
dummyhead.next = merge(dummyhead.next, lists[i])
return dummyhead.next
分治合并
class Solution:
def mergeKLists(self, lists: List[ListNode]) -> ListNode:
# 如果输入为空,直接返回空
if not lists:
return
# 获取链表列表的长度
n = len(lists)
# 调用递归函数进行合并
return self.merge(lists, 0, n-1)
def merge(self, lists, left, right):
# 当左右指针相等时,表示只有一个链表,直接返回该链表
if left == right:
return lists[left]
# 计算中间位置
mid = left + (right - left) // 2
# 递归地合并左半部分和右半部分的链表
l1 = self.merge(lists, left, mid)
l2 = self.merge(lists, mid+1, right)
# 调用合并两个有序链表的函数
return self.mergeTwoLists(l1, l2)
def mergeTwoLists(self, l1, l2):
# 若其中一个链表为空,则直接返回另一个链表
if not l1:
return l2
if not l2:
return l1
# 比较两个链表头结点的大小,选择较小的作为新链表的头结点
if l1.val < l2.val:
l1.next = self.mergeTwoLists(l1.next, l2)
return l1
else:
l2.next = self.mergeTwoLists(l1, l2.next)
return l2
最小堆
"""
假设有3个有序链表分别是:1->4->5, 1->3->4, 2->6。
初始时,最小堆为空。我们依次将(1,0),(1,1),(2,2)加入最小堆。
然后不断弹出最小值(1,0),(1,1),(2,2),加入到结果链表中,
并将对应链表的下一个节点值和索引加入最小堆,直到最小堆为空。
最终得到的合并后的链表为1->1->2->3->4->4->5->6
"""
class Solution:
def mergeKLists(self, lists: List[ListNode]) -> ListNode:
import heapq
# 创建虚拟节点
dummy = ListNode(0)
p = dummy
head = []
# 遍历链表数组
for i in range(len(lists)):
if lists[i] :
# 将每个链表的头结点值和索引加入到最小堆中
heapq.heappush(head, (lists[i].val, i))
lists[i] = lists[i].next
while head:
# 弹出最小堆中的值和对应的索引
val, idx = heapq.heappop(head)
# 创建新节点并连接到结果链表上
p.next = ListNode(val)
p = p.next
# 如果该链表还有剩余节点,则将下一个节点的值和索引加入到最小堆中
if lists[idx]:
heapq.heappush(head, (lists[idx].val, idx))
lists[idx] = lists[idx].next
# 返回合并后的链表
return dummy.next
146. LRU 缓存 - 力扣(LeetCode)
哈希 + 双向链表
借用灵神题解的图,really good
class Node:
# 提高访问属性的速度,并节省内存
__slots__ = 'prev', 'next', 'key', 'value'
def __init__(self, key=0, value=0):
# self.prev = None
# self.next = None
self.key = key
self.value = value
class LRUCache:
def __init__(self, capacity: int):
self.capacity = capacity
self.dummy = Node() # 哨兵节点
self.dummy.prev = self.dummy
self.dummy.next = self.dummy
self.key_to_node = dict()
def get_node(self, key: int) -> Optional[Node]:
if key not in self.key_to_node: # 没有这本书
return None
node = self.key_to_node[key] # 有这本书
self.remove(node) # 把这本书抽出来
self.push_front(node) # 放在最上面
return node
def get(self, key: int) -> int:
node = self.get_node(key)
return node.value if node else -1
def put(self, key: int, value: int) -> None:
node = self.get_node(key)
if node: # 有这本书
node.value = value # 更新 value
return
self.key_to_node[key] = node = Node(key, value) # 新书
self.push_front(node) # 放在最上面
if len(self.key_to_node) > self.capacity: # 书太多了
back_node = self.dummy.prev
del self.key_to_node[back_node.key]
self.remove(back_node) # 去掉最后一本书
# 删除一个节点(抽出一本书)
def remove(self, x: Node) -> None:
x.prev.next = x.next
x.next.prev = x.prev
# 在链表头添加一个节点(把一本书放在最上面)
def push_front(self, x: Node) -> None:
x.prev = self.dummy
x.next = self.dummy.next
x.prev.next = x
x.next.prev = x
# Your LRUCache object will be instantiated and called as such:
# obj = LRUCache(capacity)
# param_1 = obj.get(key)
# obj.put(key,value)
逻辑回归处理多分类 1、背景描述2、One vs One 1、背景描述 逻辑回归本身只能用于二分类问题,如果实际情况是多分类的,那么就需要对模型进行一些改动。下面介绍三种常用的将逻辑回归用于多分类的方法
2、One vs One OvO(One vs Oneÿ…
d8.bat因找不到java_exe文件,触发EOF
d8.bat中之前代码为:
set java_exe
if exist "%~dp0..\tools\lib\find_java.bat" call "%~dp0..\tools\lib\find_java.bat"
if exist "%~dp0..\..\tools\lib\find_java.bat" …