异构图中节点的分类/回归

异构图中节点的分类/回归

导入包

import numpy as np
import torch
import dgl
import torch.nn as nn
import torch.nn.functional as F
import dgl.nn as dglnn

创建一个异构图

设置这个图中的节点个数和边的个数

n_users = 100   #user节点个数
n_jobspre = 500     #jobpre节点的个数
n_uj = 3000     #边'uj'的个数
n_ju = 3000    #边'ju'的个数
n_college = 100     #college节点个数
n_cu = 100
n_uc = 100
n_hetero_features = 20     #相当于要做的嵌入维度
n_user_classes = 20 #假设得出的job种类有这么多
n_skills = 500       #skill节点个数
n_us = 4000
n_su = 4000
n_js = 4000
n_sj = 4000

设置每条边对应的头和尾（如果有数据集，则直接导入就好）这里自动生成 还不是因为没有数据集

uj_src = np.random.randint(0,n_users,n_uj)       #user_jobpre中user的编号
uj_dst = np.random.randint(0,n_jobspre,n_uj)     #user_jobpre中jobpre的编号
uc_src = np.random.randint(0,n_users,n_cu)       #user_college中user的编号
uc_dst = np.random.randint(0,n_college,n_cu)     #user_college中college的编号
us_src = np.random.randint(0,n_users,n_us)      #user_skill中user的编号
us_dst = np.random.randint(0,n_skills,n_us)       #user_skill中skill的编号
js_src = np.random.randint(0,n_jobspre,n_js)
js_dst = np.random.randint(0,n_skills,n_js)

利用dgl构建异构图

hetero_graph = dgl.heterograph({
    ('user', 'uj', 'jobpre'): (uj_src, uj_dst),
    ('jobpre', 'ju', 'user'): (uj_dst, uj_src),
    ('user', 'uc', 'college'): (uc_src, uc_dst),
    ('college', 'cu', 'user'): (uc_dst, uc_src),
    ('user', 'us', 'skill'): (us_src, us_dst),
    ('skill', 'su', 'user'): (us_dst, us_src),
    ('jobpre','js','skill'):(js_src,js_dst),
    ('skill','sj','jobpre'):(js_dst,js_src),
    })

初始化每个节点的嵌入

hetero_graph.nodes['user'].data['feature'] = torch.randn(n_users, n_hetero_features)  #给user node 添加属性，相当于用户嵌入维度为20
hetero_graph.nodes['jobpre'].data['feature'] = torch.randn(n_jobspre, n_hetero_features) 
hetero_graph.nodes['college'].data['feature'] = torch.randn(n_college, n_hetero_features)
hetero_graph.nodes['skill'].data['feature'] = torch.randn(n_skills, n_hetero_features)
hetero_graph.nodes['user'].data['label'] = torch.randint(0, n_user_classes, (n_users,))    #用户的个人标签
hetero_graph.nodes['user'].data['train_mask'] = torch.zeros(n_users, dtype=torch.bool).bernoulli(0.6) #选出一些计算损失函数

定义一个异构卷积层

class RGCN(nn.Module):
    def __init__(self, in_feats, hid_feats, out_feats, rel_names):
        super().__init__()
        # 实例化HeteroGraphConv，in_feats是输入特征的维度，out_feats是输出特征的维度，aggregate是聚合函数的类型
        self.conv1 = dglnn.HeteroGraphConv({
            rel: dglnn.GraphConv(in_feats, hid_feats)
            for rel in rel_names}, aggregate='sum')
        self.conv2 = dglnn.HeteroGraphConv({
            rel: dglnn.GraphConv(hid_feats, out_feats)
            for rel in rel_names}, aggregate='sum')

    def forward(self, graph, inputs):   #inputs： node_features
        # 输入是节点的特征字典
        h = self.conv1(graph, inputs)
        h = {k: F.relu(v) for k, v in h.items()}
        h = self.conv2(graph, h)
        return h

定义模型并进行训练

model = RGCN(n_hetero_features, 20, n_user_classes, hetero_graph.etypes)       #['clicked-by', 'disliked-by', 'click', 'dislike', 'follow', 'followed-by']
user_feats = hetero_graph.nodes['user'].data['feature']         #用户的特征嵌入
jobpre_feats = hetero_graph.nodes['jobpre'].data['feature']      #物品的特征嵌入
college_feats = hetero_graph.nodes['college'].data['feature']      #物品的特征嵌入
skill_feats = hetero_graph.nodes['skill'].data['feature']      #物品的特征嵌入
labels = hetero_graph.nodes['user'].data['label']
train_mask = hetero_graph.nodes['user'].data['train_mask']

node_features = {'user': user_feats, 'jobpre': jobpre_feats,'college':college_feats,'skill':skill_feats}        #所有用户的特征嵌入     所有物品的特征嵌入
opt = torch.optim.Adam(model.parameters())

for epoch in range(5):
    model.train()
    # 使用所有节点的特征进行前向传播计算，并提取输出的user节点嵌入
    logits = model(hetero_graph, node_features)['user']
    h_dict = model(hetero_graph, {'user': user_feats, 'jobpre': jobpre_feats,'college':college_feats,'skill':skill_feats} )
    h_user = h_dict['user']     #模型每次迭代后得出的 用户的特征嵌入
    print(h_user[2])  # 第2个用户的特征，输出特征为5，因为用户的分类为5个
    h_jobpre= h_dict['jobpre']       #模型每次迭代后得出的 物品的特征嵌入
    h_college = h_dict['college']
    h_skill = h_dict['skill']
    
    # 计算损失值
    loss = F.cross_entropy(logits[train_mask], labels[train_mask])
    # 计算验证集的准确度。在本例中省略。
    # 进行反向传播计算
    opt.zero_grad()
    loss.backward()
    opt.step()
    print(loss.item())
    torch.save(model.state_dict(), "./model/main01" + "_" + str(epoch))   #保存模型

使用（测试）模型

model = RGCN(n_hetero_features, 20, n_user_classes, hetero_graph.etypes)
model.load_state_dict(torch.load("./model/main01_4" ))

user_feats = hetero_graph.nodes['user'].data['feature']
print(user_feats )
print(user_feats[0])

结果截图：