def Search(self):
profiles = set()
# 随机加一个点进入
# seed = random.randint(0,self.num)
# 加入相邻最多的一个点进来
seed = random.choice(self.id_list)
# results = {i:len(set(all_neighbors(self.g,str(i)))) for i in self.id_list if i in self.g}
# seed = max(results.items(),key=lambda key:key[1])[0]
# neighbours = set([i for i in self.Neighbours(seed)])
neighbours = set([
i for i in (all_neighbors(self.g, str(seed)) if str(seed) in
self.g else set())
])
profiles.add(seed)
neighbours.add(seed)
while len(profiles) < self.k:
# 每次加入使得|N(v) - (N(S) | S)|最大的元素
# results = {i:len(set((all_neighbors(self.g,i) if i in self.g else set())) - profiles - neighbours ) for i in range(self.num) if i not in profiles}
results = {
i: len(
set((all_neighbors(self.g, str(i)) if str(i) in
self.g else set())) - profiles - neighbours)
for i in self.id_list if i not in profiles
}
to_add = max(results.items(), key=lambda key: key[1])[0]
profiles.add(to_add)
# 更新neighbours
neighbours |= set([
i for i in (all_neighbors(self.g, to_add) if to_add in
self.g else set())
])
neighbours.add(to_add)
# print len(profiles)
profiles = [int(i) for i in profiles]
return profiles
def Search(self):
profiles = set()
# 随机加一个点进入
seed = random.randint(0, self.num)
# neighbours = set([i for i in self.Neighbours(seed)])
neighbours = set([
i
for i in (all_neighbors(self.g, seed) if seed in self.g else set())
])
while len(profiles) < self.k:
# 每次加入使得|N(v) - (N(S) | S)|最大的元素
results = {
i: len(
set((all_neighbors(self.g, i) if i in self.g else set())) -
profiles - neighbours | profiles)
for i in range(self.num) if i not in profiles
}
to_add = max(results.items(), key=lambda key: key[1])[0]
profiles.add(to_add)
# 更新neighbours
neighbours |= set([
i for i in (all_neighbors(self.g, to_add) if to_add in
self.g else set())
])
print len(profiles)
profiles = [int(self.id_list[i]) for i in profiles]
return profiles
def Rt(self, profiles):
neighbours = reduce(lambda x, y: x | y, [
set(all_neighbors(self.g, self.id_dic[str(u)]))
if self.id_dic[str(u)] in self.g else set() for u in profiles
])
NS = len(neighbours - set(profiles))
VS = self.num - len(profiles)
return 1.0 * NS / VS
def GetNeighbours(self, u):
if self.id_dic[str(u)] in self.g:
return set([
int(self.id_list[i])
for i in all_neighbors(self.g, self.id_dic[str(u)])
])
else:
return set()
def Sample(self, gamma, clusters, id_cluster):
num = len(self.users)
profiles = set()
# 初始化每个cluster的Rc和Rt
Rc = {i: 0 for i in clusters.keys()}
Rt = {i: 0 for i in clusters.keys()}
neighbours = {i: set() for i in id_cluster.keys()}
while len(profiles) < self.need:
results = {}
# 对于不在profiles中的元素继续加入,加入使得F最大的元素
# 从每个聚类簇中选取元素加入
for key in clusters.keys():
temp = {
i: len(clusters[key]) * 1.0 / num *
(gamma * self.CalcRc(profiles | set([i]), clusters[key]) -
gamma * Rc[key] + (1 - gamma) * (len(
set(
all_neighbors(self.g, self.users.iloc[i]
['userid'])) & id_cluster[key] -
neighbours[key]) if self.users.iloc[i]['userid']
in self.g else 0))
for i in clusters[key] if i not in profiles
}
# 找到temp中增长最大的值和结果
item = max(temp.items(), key=lambda key: key[1])
results[item[0]] = item[1]
# 从results找到增长最大的加入
to_add = max(results.items(), key=lambda key: key[1])[0]
# to_add = results.keys()[to_add]
profiles.add(to_add)
# 确定to_add属于哪个聚类簇
for key in clusters.keys():
if to_add in clusters[key]:
belong = key
break
print len(profiles)
# 更新belong的Rc和Rt,neigbours
neighbours[belong] |= set(
all_neighbors(self.g, self.users.iloc[to_add]['userid'])
) if self.users.iloc[to_add]['userid'] in self.g else set()
Rc[belong] = self.CalcRc(profiles, clusters[belong])
Rt[belong] = self.CalcRt(profiles, clusters[belong])
profiles = [self.users.iloc[i]['userid'] for i in profiles]
return profiles
def Rt(self, users, profiles):
# neighbours = reduce(lambda x,y:x | y,[set(all_neighbors(self.g,self.id_dic[str(u)])) if self.id_dic[str(u)] in self.g else set() for u in profiles])
neighbours = reduce(lambda x, y: x | y, [
set(all_neighbors(self.g, str(u))) if str(u) in self.g else set()
for u in profiles
])
# 所有相邻的
NS = len(neighbours - set(profiles) & set(users))
VS = len(users)
return 1.0 * NS / VS
def CalcRt(self, profiles, cluster):
'''
:param profiles: 代表性子集
:param cluster: 某个聚类簇
:return:
'''
clusteri = reduce(
lambda x, y: x | y,
[set([self.users.iloc[i]['userid']]) for i in cluster])
neighbours = reduce(lambda x, y: x | y, [
set(all_neighbors(self.g, str(self.users.iloc[u]['userid'])))
if str(u) in self.g else set() for u in profiles
])
return len(neighbours & clusteri) * 1.0 / len(clusteri)
def GetNeighbours(self, u):
if str(u) in self.g:
return set([int(i) for i in all_neighbors(self.g, str(u))])
else:
return set()