最小生成树----Kruskal算法&并查集

对于图寻找最小生成树

顶点v使用并查集，初始状态下所有顶点都是一个集合。

Kruskal算法描述：

取最短的边，比较是否两个顶点都在同一集合内：如果都在，则舍弃

初始化时，v长这样

v = [i for i in xrange(0, length)] 

并(i, j)的操作就是合并两个集合的跟：

r_i = find_root(v, i)
r_j = find_root(v, j)
v[r_j] = r_i

查找操作中，我增加了路径压缩，使得效率能提升

def find_root(d_set, key):
	if d_set[key] == key:
		return key
	# flat
	root = key
	cur_key = key
	while d_set[cur_key] != cur_key:
		root = d_set[cur_key]
		cur_key = root
	d_set[key] = root
	return root

最后是kruskal的算法了：

def kruskal(graph):
	# only traverse half of the graph
	# because graph is undirected
	length = len(graph)
	# init
	edge_info = []          # (weight, i, j) list
	for i in xrange(0, len(graph)):
		for j in xrange(0, i):
			if graph[i][j]:
				edge_info.append((graph[i][j], i, j))
	edge_info.sort()

	v = [i for i in xrange(0, length)]     # disjoint set
	e = []
	for (_, i, j) in edge_info:
		r_i = find_root(v, i)
		r_j = find_root(v, j)
		if r_i == r_j:
			continue
		# union
		v[r_j] = r_i
		e.append((i, j))
	return e

驱动

graph = [[0, 7, 9,  0,  0, 14],
         [7, 0, 10, 15, 0, 0],
         [9, 10, 0, 11, 0, 2],
         [0, 15, 11, 0, 6, 0],
         [0, 0,  0,  6, 0, 9],
         [14, 0, 2, 0,  9, 0]]
print kruskal(graph)

排序需要O(Edge * log(Edge))，查找时是O(Edge * 接近常数的数)

最终时间复杂度为O(Edge * log(Edge))