add: homework11

This commit is contained in:
SunnyQjm committed 2020-06-02 19:33:05 +08:00
1 parent 351cc98168
commit 9b0742e93c
5 files changed
+96 -1

No files matched your search

Binary file not shown.

After

Width:  |  Height:  |  Size: 19 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 19 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 19 KiB

+95
View File
@@ -0,0 +1,95 @@
# Algorithm Homework 11
> 姓名:阙建明
>
> 学号:1901213139
## 作业要求:
Leetcode: 743, 847
- ### Leetcode problem 743
> [网络延迟时间](https://leetcode-cn.com/problems/network-delay-time/)
- **算法思路:**
1. 使用Dijkstra算法找到发信号的点到其它所有点的最短路径(时间作为边的权重);
2. 结束时判断是否遍历了所有的点,没有则返回-1,有则返回所有点里面距离最大的点的距离作为总的耗时。
3. 遇到大于限定跳数就跳出进行剪枝;
4. 遇到目标节点就直接返回其最小开销,结束算法。
- **Leetcode提交结果:**
![image-20200602172430250](AlgorithmHomework11.assets/image-20200602172430250.png)
- **代码:**
```python
class Solution:
def networkDelayTime(self, times: List[List[int]], N: int, K: int) -> int:
# 构造图
graph = collections.defaultdict(list)
for u, v, w in times:
graph[u].append((w, v))
# 用一个堆来存储已只一个非无穷的距离,但是还没用进行松弛的item => 每个item包含 当前已知的源到点的距离,以及点的编号
pq = [(0, K)]
# 记录已经松弛的点
visited = {}
while pq:
distance, node = heapq.heappop(pq)
# 这边过滤到多余push的点
if node in visited:
continue
visited[node] = distance
for distance2, neighbour in graph[node]:
if neighbour not in visited:
# 遍历每个邻居的时候都会执行push操作,可能有一个点被push若干次,但是由于堆的特性,最先被访问到的一定是距离最小的那个
heapq.heappush(pq, (distance + distance2, neighbour))
return max(visited.values()) if len(visited) == N else -1
```
- ### Leetcode problem 934
> [最短的桥](https://leetcode-cn.com/problems/shortest-bridge/)
- **算法思路:**
1. 以每个点为起点,同时进行BFS搜索;
2. 用 `(cover, cur)` 元组代表一个状态 (State)=> 其中cover表示当前已经走过的节点(用二进制位的压缩表示,每个二进制位代表一个节点是否走过),cur表示当前所出的节点;
3. 用dict字典记录达成每个状态(State)所需的最小跳数;
4. 每次从队列中取出一个状态,遍历当前节点的所有邻居转移到若干个新状态,添加到队列当中(同时使用dict来过滤掉一些不必要的状态)。
- **Leetcode提交结果:**
![image-20200602192555935](AlgorithmHomework11.assets/image-20200602192555935.png)
- **代码:**
```python
class Solution:
def shortestPathLength(self, graph: List[List[int]]) -> int:
N = len(graph)
# 使用队列,保证先进先出,可以用来做BFS遍历
queue = collections.deque((1 << i, i) for i in range(N))
# 最短路径必然小于N * N
dist = collections.defaultdict(lambda: N * N)
# 结束标记
END = (1 << N) - 1
# 初始化,出发点所走路径为0
for i in range(N):
dist[(1 << i, i)] = 0
while queue:
cover, cur = queue.popleft()
d = dist[(cover, cur)]
if cover == END:
return d
for child in graph[cur]:
newCover = cover | (1 << child)
if d + 1 < dist[newCover, child]: # 每个状态只记录最短路径
dist[newCover, child] = d + 1
queue.append((newCover, child))
```
+1 -1
View File
@@ -9,6 +9,6 @@
- [第七次作业](AlgorithmHomework7.md)
- [第八次作业](AlgorithmHomework8.md)
- [第九次作业](AlgorithmHomework9.md)
- [第十次作业](AlgorithmHomework10.md)
- [第十一次作业](AlgorithmHomework11.md)