1.算法分析
同node2vec一致,我们也需要先找到所有节点的
γ
\gamma
γ条随机游走路径,每条路径长度为
t
t
t。
DeepWalk最大的贡献是将Skip-Gram模型引入到了图嵌入中,至于随机游走方式则并不如node2vec那样复杂。
DeepWalk中的RandomWalk:每次随机选择当前节点的一个邻居节点!
2.代码实现
2.1 RandomWalk
def random_walk(self, u):
g = self.G
walk = [u]
while len(walk) < self.l:
curr = walk[-1]
v_curr = list(g.neighbors(curr))
if len(v_curr) > 0:
walk.append(random.choice(v_curr))
else:
break
return walk
代码分析:
walk.append(random.choice(v_curr))
每次随机选择当前节点的一个邻居节点作为下一个节点。
2.2 Skip-Gram
得到所有节点的全部路径后,利用gensim库中的Word2Vec来进行训练:
def learning_features(self):
g = self.G
walks = []
nodes = list(g.nodes())
for t in range(self.r):
random.shuffle(nodes)
for node in nodes:
walk = self.random_walk(node)
walks.append(walk)
# embedding
walks = [list(map(str, walk)) for walk in walks]
model = Word2Vec(sentences=walks, vector_size=self.d, window=self.k, min_count=0, sg=1, workers=3)
f = model.wv
return f
调用:
if __name__ == '__main__':
d, r, l, k = 128, 10, 80, 10
G = nx.les_miserables_graph()
deep_walk = deep_walk(G, d, r, l, k)
model = deep_walk.learning_features()
3.节点聚类分析
得到所有节点的向量表示后,简单利用k-means对所有节点进行聚类:
def k_means(self, m, K):
"""
:param m: deepwalk的训练结果
:param K: 类别数
:return: 聚类结果
"""
d = self.d
nodes = list(G.nodes)
# 任意选择K个节点作为初始聚类中心
centers = []
temp = []
for i in range(K):
t = np.random.randint(0, len(nodes) - 1)
if nodes[t] not in centers:
temp.append(nodes[t])
centers.append(m[nodes[t]]) # 中心为向量
# 迭代50次
res = {}
for i in range(K):
res[i] = []
for time in range(50):
# clear
for i in range(K):
res[i].clear()
# 算出每个点的向量到聚类中心的距离
nodes_distance = {}
for node in nodes:
# node到中心节点的距离
node_distance = []
for center in centers:
node_distance.append(get_dis(m[node], center))
nodes_distance[node] = node_distance # 保存node节点到各个中心的距离
# 对每个节点重新划分类别,选择一个最近的节点进行分类,类别为0-5
for node in nodes:
temp = nodes_distance[node] # 存放着6个距离
cls = temp.index(min(temp))
res[cls].append(node)
# 更新聚类中心
centers.clear()
for i in range(K):
center = []
for j in range(d):
t = [m[node][j] for node in res[i]] # 第i个类别中所有node节点的第j个坐标
center.append(np.mean(t))
centers.append(center)
return res
画图:
def plot(self, m, K):
"""
:param m: 节点的向量表示
:param K: 聚类个数
:return: none
"""
g = self.G
# 根据原始标签画图
pos = nx.spring_layout(G)
color_map = []
ns = list(G.nodes.data())
nodes = list(g.nodes)
# for node in range(len(nodes)):
# if ns[node][1]['bipartite'] == 0:
# color_map.append('#DCBB8A')
# else:
# color_map.append('#98BBEF')
#
# plt.subplot(2, 1, 1)
# nx.draw(G, node_color=color_map, pos=pos, with_labels=True, node_size=1000)
res = self.k_means(m, K)
colors = ['#DCBB8A', '#98BBEF', 'navy', 'indigo', 'orange', 'blue']
color_map.clear()
for node in nodes:
for i in range(len(res)):
if node in res[i]:
color_map.append(colors[i])
break
# draw
# plt.subplot(2, 1, 2)
nx.draw(G, node_color=color_map, pos=pos, with_labels=True, node_size=1000)
plt.show()
node2vec中对《悲惨世界》中的任务关系做了节点聚类分析,共分为六类,结果如下:
采用DeepWalk的效果为(k-means迭代50次):
迭代100次: