Press "Enter" to skip to content

Python实现基于图神经网络的异构图表示学习和推荐算法研究

本站内容均来自兴趣收集,如不慎侵害的您的相关权益,请留言告知,我们将尽快删除.谢谢.

资源下载地址:https://download.csdn.net/download/sheziqiong/85978304

 

资源下载地址:https://download.csdn.net/download/sheziqiong/85978304

 

目录结构

 

GNN-Recommendation/
    gnnrec/             算法模块顶级包
        hge/            异构图表示学习模块
        kgrec/          基于图神经网络的推荐算法模块
    data/               数据集目录(已添加.gitignore)
    model/              模型保存目录(已添加.gitignore)
    img/                图片目录
    academic_graph/     Django项目模块
    rank/               Django应用
    manage.py           Django管理脚本

 

安装依赖

 

Python 3.7

 

CUDA 11.0

 

pip install -r requirements_cuda.txt

 

pip install -r requirements.txt

 

异构图表示学习(附录)

 

基于对比学习的关系感知异构图神经网络(Relation-aware Heterogeneous Graph Neural Network with Contrastive Learning, RHCO)

 

 

readme

 

基于图神经网络的推荐算法(附录)

 

基于图神经网络的学术推荐算法(Graph Neural Network based Academic Recommendation Algorithm, GARec)

 

 

readme

 

Django 配置

 

MySQL 数据库配置

 

 

    1. 创建数据库及用户

 

 

CREATE DATABASE academic_graph CHARACTER SET utf8mb4;
CREATE USER 'academic_graph'@'%' IDENTIFIED BY 'password';
GRANT ALL ON academic_graph.* TO 'academic_graph'@'%';

 

 

    1. 在根目录下创建文件.mylogin.cnf

 

 

[client]
host = x.x.x.x
port = 3306
user = username
password = password
database = database
default-character-set = utf8mb4

 

 

    1. 创建数据库表

 

 

python manage.py makemigrations --settings=academic_graph.settings.prod rank
python manage.py migrate --settings=academic_graph.settings.prod

 

 

    1. 导入 oag-cs 数据集

 

 

python manage.py loadoagcs --settings=academic_graph.settings.prod

 

注:由于导入一次时间很长(约 9 小时),为了避免中途发生错误,可以先用 data/oag/test 中的测试数据调试一下

 

拷贝静态文件

 

python manage.py collectstatic --settings=academic_graph.settings.prod

 

启动 Web 服务器

 

export SECRET_KEY=xxx
python manage.py runserver --settings=academic_graph.settings.prod 0.0.0.0:8000

 

系统截图

 

搜索论文

 

论文详情

 

搜索学者

 

学者详情

 

基于图神经网络的推荐算法

 

oag-cs – 使用 OAG 微软学术数据构造的计算机领域的学术网络(见readme)

 

预训练顶点嵌入

 

使用 metapath2vec(随机游走 +word2vec)预训练顶点嵌入,作为 GNN 模型的顶点输入特征

 

 

    1. 随机游走

 

 

python -m gnnrec.kgrec.random_walk model/word2vec/oag_cs_corpus.txt

 

 

    1. 训练词向量

 

 

python -m gnnrec.hge.metapath2vec.train_word2vec --size=128 --workers=8 model/word2vec/oag_cs_corpus.txt model/word2vec/oag_cs.model

 

使用微调后的 SciBERT 模型(见readme 第 2 步)将查询词编码为向量,与预先计算好的论文标题向量计算余弦相似度,取 top k

 

python -m gnnrec.kgrec.recall

 

召回结果示例:

 

graph neural network

 

0.9629Aggregation Graph Neural Networks
0.9579Neural Graph Learning: Training Neural Networks Using Graphs
0.9556Heterogeneous Graph Neural Network
0.9552Neural Graph Machines: Learning Neural Networks Using Graphs
0.9490On the choice of graph neural network architectures
0.9474Measuring and Improving the Use of Graph Information in Graph Neural Networks
0.9362Challenging the generalization capabilities of Graph Neural Networks for network modeling
0.9295Strategies for Pre-training Graph Neural Networks
0.9142Supervised Neural Network Models for Processing Graphs
0.9112Geometrically Principled Connections in Graph Neural Networks

 

recommendation algorithm based on knowledge graph

 

0.9172Research on Video Recommendation Algorithm Based on Knowledge Reasoning of Knowledge Graph
0.8972An Improved Recommendation Algorithm in Knowledge Network
0.8558A personalized recommendation algorithm based on interest graph
0.8431An Improved Recommendation Algorithm Based on Graph Model
0.8334The Research of Recommendation Algorithm based on Complete Tripartite Graph Model
0.8220Recommendation Algorithm based on Link Prediction and Domain Knowledge in Retail Transactions
0.8167Recommendation Algorithm Based on Graph-Model Considering User Background Information
0.8034A Tripartite Graph Recommendation Algorithm Based on Item Information and User Preference
0.7774Improvement of TF-IDF Algorithm Based on Knowledge Graph
0.7770Graph Searching Algorithms for Semantic-Social Recommendation

 

scholar disambiguation

 

0.9690Scholar search-oriented author disambiguation
0.9040Author name disambiguation in scientific collaboration and mobility cases
0.8901Exploring author name disambiguation on PubMed-scale
0.8852Author Name Disambiguation in Heterogeneous Academic Networks
0.8797KDD Cup 2013: author disambiguation
0.8796A survey of author name disambiguation techniques: 2010–2016
0.8721Who is Who: Name Disambiguation in Large-Scale Scientific Literature
0.8660Use of ResearchGate and Google CSE for author name disambiguation
0.8643Automatic Methods for Disambiguating Author Names in Bibliographic Data Repositories
0.8641A brief survey of automatic methods for author name disambiguation

 

构造 ground truth

 

(1)验证集

 

从 AMiner 发布的 AI 2000 人工智能全球最具影响力学者榜单 抓取人工智能 20 个子领域的 top 100 学者

 

pip install scrapy>=2.3.0
cd gnnrec/kgrec/data/preprocess
scrapy runspider ai2000_crawler.py -a save_path=/home/zzy/GNN-Recommendation/data/rank/ai2000.json

 

与 oag-cs 数据集的学者匹配,并人工确认一些排名较高但未匹配上的学者,作为学者排名 ground truth 验证集

 

export DJANGO_SETTINGS_MODULE=academic_graph.settings.common
export SECRET_KEY=xxx
python -m gnnrec.kgrec.data.preprocess.build_author_rank build-val

 

(2)训练集

 

参考 AI 2000 的计算公式,根据某个领域的论文引用数加权求和构造学者排名,作为 ground truth 训练集

 

计算公式:

 

即:假设一篇论文有 n 个作者,第 k 作者的权重为 1/k,最后一个视为通讯作者,权重为 1/2,归一化之后计算论文引用数的加权求和

 

python -m gnnrec.kgrec.data.preprocess.build_author_rank build-train

 

(3)评估 ground truth 训练集的质量

 

python -m gnnrec.kgrec.data.preprocess.build_author_rank eval

 

[email protected]=0.2420 [email protected]=0.1859    [email protected]=0.2016
[email protected]=0.2308  [email protected]=0.2494     [email protected]=0.1351
[email protected]=0.2492  [email protected]=0.3118     [email protected]=0.0678
[email protected]=0.2743  [email protected]=0.3471     [email protected]=0.0376
[email protected]=0.3165   Pr[email protected]=0.3765      [email protected]=0.0203

 

(4)采样三元组

 

从学者排名训练集中采样三元组(t, ap, an),表示对于领域 t,学者 ap 的排名在 an 之前

 

python -m gnnrec.kgrec.data.preprocess.build_author_rank sample

 

训练 GNN 模型

 

python -m gnnrec.kgrec.train model/word2vec/oag-cs.model model/garec_gnn.pt data/rank/author_embed.pt

 

异构图表示学习

ACM – ACM 学术网络数据集
DBLP – DBLP 学术网络数据集
ogbn-mag – OGB 提供的微软学术数据集
oag-venue – oag-cs 期刊分类数据集

数据集顶点数边数目标顶点类别数
ACM1124634852paper3
DBLP26128239566author4
ogbn-mag193974321111007paper349
oag-venue423516934520417paper360

 

Baselines

R-GCN
HGT
HGConv
R-HGNN
C&S
HeCo

R-GCN (full batch)

 

python -m gnnrec.hge.rgcn.train --dataset=acm --epochs=10
python -m gnnrec.hge.rgcn.train --dataset=dblp --epochs=10
python -m gnnrec.hge.rgcn.train --dataset=ogbn-mag --num-hidden=48
python -m gnnrec.hge.rgcn.train --dataset=oag-venue --num-hidden=48 --epochs=30

 

(使用 minibatch 训练准确率就是只有 20% 多,不知道为什幺)

 

预训练顶点嵌入

 

使用 metapath2vec(随机游走 +word2vec)预训练顶点嵌入,作为 GNN 模型的顶点输入特征

 

python -m gnnrec.hge.metapath2vec.random_walk model/word2vec/ogbn-mag_corpus.txt
python -m gnnrec.hge.metapath2vec.train_word2vec --size=128 --workers=8 model/word2vec/ogbn-mag_corpus.txt model/word2vec/ogbn-mag.model

 

python -m gnnrec.hge.hgt.train_full --dataset=acm
python -m gnnrec.hge.hgt.train_full --dataset=dblp
python -m gnnrec.hge.hgt.train --dataset=ogbn-mag --node-embed-path=model/word2vec/ogbn-mag.model --epochs=40
python -m gnnrec.hge.hgt.train --dataset=oag-venue --node-embed-path=model/word2vec/oag-cs.model --epochs=40

 

python -m gnnrec.hge.hgconv.train_full --dataset=acm --epochs=5
python -m gnnrec.hge.hgconv.train_full --dataset=dblp --epochs=20
python -m gnnrec.hge.hgconv.train --dataset=ogbn-mag --node-embed-path=model/word2vec/ogbn-mag.model
python -m gnnrec.hge.hgconv.train --dataset=oag-venue --node-embed-path=model/word2vec/oag-cs.model

 

python -m gnnrec.hge.rhgnn.train_full --dataset=acm --num-layers=1 --epochs=15
python -m gnnrec.hge.rhgnn.train_full --dataset=dblp --epochs=20
python -m gnnrec.hge.rhgnn.train --dataset=ogbn-mag model/word2vec/ogbn-mag.model
python -m gnnrec.hge.rhgnn.train --dataset=oag-venue --epochs=50 model/word2vec/oag-cs.model

 

python -m gnnrec.hge.cs.train --dataset=acm --epochs=5
python -m gnnrec.hge.cs.train --dataset=dblp --epochs=5
python -m gnnrec.hge.cs.train --dataset=ogbn-mag --prop-graph=data/graph/pos_graph_ogbn-mag_t5.bin
python -m gnnrec.hge.cs.train --dataset=oag-venue --prop-graph=data/graph/pos_graph_oag-venue_t5.bin

 

python -m gnnrec.hge.heco.train --dataset=ogbn-mag model/word2vec/ogbn-mag.model data/graph/pos_graph_ogbn-mag_t5.bin
python -m gnnrec.hge.heco.train --dataset=oag-venue model/word2vec/oag-cs.model data/graph/pos_graph_oag-venue_t5.bin

 

(ACM 和 DBLP 的数据来自 https://github.com/ZZy979/pytorch-tutorial/tree/master/gnn/heco ,准确率和 Micro-F1 相等)

 

基于对比学习的关系感知异构图神经网络(Relation-aware Heterogeneous Graph Neural Network with Contrastive Learning, RHCO)

 

在 HeCo 的基础上改进:

网络结构编码器中的注意力向量改为关系的表示(类似于 R-HGNN)
正样本选择方式由元路径条数改为预训练的 HGT 计算的注意力权重、训练集使用真实标签
元路径视图编码器改为正样本图编码器,适配 mini-batch 训练
Loss 增加分类损失,训练方式由无监督改为半监督
在最后增加 C&S 后处理步骤

ACM

 

python -m gnnrec.hge.hgt.train_full --dataset=acm --save-path=model/hgt/hgt_acm.pt
python -m gnnrec.hge.rhco.build_pos_graph_full --dataset=acm --num-samples=5 --use-label model/hgt/hgt_acm.pt data/graph/pos_graph_acm_t5l.bin
python -m gnnrec.hge.rhco.train_full --dataset=acm data/graph/pos_graph_acm_t5l.bin

 

DBLP

 

python -m gnnrec.hge.hgt.train_full --dataset=dblp --save-path=model/hgt/hgt_dblp.pt
python -m gnnrec.hge.rhco.build_pos_graph_full --dataset=dblp --num-samples=5 --use-label model/hgt/hgt_dblp.pt data/graph/pos_graph_dblp_t5l.bin
python -m gnnrec.hge.rhco.train_full --dataset=dblp --use-data-pos data/graph/pos_graph_dblp_t5l.bin

 

ogbn-mag(第 3 步如果中断可使用–load-path 参数继续训练)

 

python -m gnnrec.hge.hgt.train --dataset=ogbn-mag --node-embed-path=model/word2vec/ogbn-mag.model --epochs=40 --save-path=model/hgt/hgt_ogbn-mag.pt
python -m gnnrec.hge.rhco.build_pos_graph --dataset=ogbn-mag --num-samples=5 --use-label model/word2vec/ogbn-mag.model model/hgt/hgt_ogbn-mag.pt data/graph/pos_graph_ogbn-mag_t5l.bin
python -m gnnrec.hge.rhco.train --dataset=ogbn-mag --num-hidden=64 --contrast-weight=0.9 model/word2vec/ogbn-mag.model data/graph/pos_graph_ogbn-mag_t5l.bin model/rhco_ogbn-mag_d64_a0.9_t5l.pt
python -m gnnrec.hge.rhco.smooth --dataset=ogbn-mag model/word2vec/ogbn-mag.model data/graph/pos_graph_ogbn-mag_t5l.bin model/rhco_ogbn-mag_d64_a0.9_t5l.pt

 

oag-venue

 

python -m gnnrec.hge.hgt.train --dataset=oag-venue --node-embed-path=model/word2vec/oag-cs.model --epochs=40 --save-path=model/hgt/hgt_oag-venue.pt
python -m gnnrec.hge.rhco.build_pos_graph --dataset=oag-venue --num-samples=5 --use-label model/word2vec/oag-cs.model model/hgt/hgt_oag-venue.pt data/graph/pos_graph_oag-venue_t5l.bin
python -m gnnrec.hge.rhco.train --dataset=oag-venue --num-hidden=64 --contrast-weight=0.9 model/word2vec/oag-cs.model data/graph/pos_graph_oag-venue_t5l.bin model/rhco_oag-venue.pt
python -m gnnrec.hge.rhco.smooth --dataset=oag-venue model/word2vec/oag-cs.model data/graph/pos_graph_oag-venue_t5l.bin model/rhco_oag-venue.pt

 

消融实验

 

python -m gnnrec.hge.rhco.train --dataset=ogbn-mag --model=RHCO_sc model/word2vec/ogbn-mag.model data/graph/pos_graph_ogbn-mag_t5l.bin model/rhco_sc_ogbn-mag.pt
python -m gnnrec.hge.rhco.train --dataset=ogbn-mag --model=RHCO_pg model/word2vec/ogbn-mag.model data/graph/pos_graph_ogbn-mag_t5l.bin model/rhco_pg_ogbn-mag.pt

 

实验结果

 

顶点分类

 

参数敏感性分析

 

消融实验

 

资源下载地址:https://download.csdn.net/download/sheziqiong/85978304

 

资源下载地址:https://download.csdn.net/download/sheziqiong/85978304

Be First to Comment

发表评论

您的电子邮箱地址不会被公开。