当前位置： 首页 > web >正文

GCN图卷积神经网络的Pytorch实现

web 2025/8/19 21:12:25

一、Pytorch代码

使用Cora数据集，采用两次GCN卷积操作，进行节点分类任务。

import torch
import torch.nn.functional as F
from torch_geometric.graphgym.register import optimizer_dict
from torch_geometric.nn import GCNConv,global_mean_pool
from torch_geometric.datasets import Planetoid
from torch_geometric.data import DataLoader
import matplotlib.pyplot as plt
plt.rcParams['font.sans-serif'] = ['SimHei']  # 设置中文字体
plt.rcParams['axes.unicode_minus'] = False   # 正常显示负号
import osclass Net(torch.nn.Module):def __init__(self):super().__init__()self.conv1=GCNConv(dataset.num_node_features,16)self.conv2=GCNConv(16,dataset.num_classes)def forward(self,data):x,edge_index=data.x,data.edge_indexx=self.conv1(x,edge_index)x=F.relu(x)x=F.dropout(x,training=self.training)x=self.conv2(x,edge_index)return F.log_softmax(x,dim=1)if  __name__=="__main__":dataset = Planetoid(root="./data/Cora", name="Cora")data=dataset[0]model=Net()optimizer=torch.optim.Adam(model.parameters(),lr=0.01,weight_decay=5e-4)Loss=[]test_acc=[]model.train()for epoch in range(200):optimizer.zero_grad()out=model(data)loss=F.nll_loss(out[data.train_mask],data.y[data.train_mask])   #负对数似然损失函数loss.backward()optimizer.step()Loss.append(loss.item())model.eval()_,pred=model(data).max(dim=1)correct=int(pred[data.test_mask].eq(data.y[data.test_mask]).sum().item())acc=correct/int(data.test_mask.sum())test_acc.append(acc)print(f"Epoch:{epoch}   Train Loss:{loss.item()}   Test Acc:{acc}")base_dir="C:\\Users\\Administrator\\PycharmProjects\\PythonProject2\\PythonProject\\PythonProject\\PythonProject\\GNN"dir1=os.path.join(base_dir,"cora_gcn_loss.jpg")dir2=os.path.join(base_dir,"cora_gcn_test_acc.jpg")plt.figure(figsize=(12,6))plt.plot(Loss,color="tomato")plt.xlabel("epoch")plt.ylabel("loss")plt.title("训练集的损失函数图")plt.grid()plt.savefig(dir1,dpi=300)plt.close()plt.figure(figsize=(12,6))plt.plot(test_acc,color="orange")plt.xlabel("epoch")plt.ylabel("acc")plt.title("测试集的准确率")plt.grid()plt.savefig(dir2,dpi=300)plt.close()

二、运行结果

2.1 损失函数图

2.2 测试集准确率变化

2.3 每个epoch的输出

Epoch:0   Train Loss:1.9429240226745605   Test Acc:0.473
Epoch:1   Train Loss:1.8335933685302734   Test Acc:0.593
Epoch:2   Train Loss:1.6950715780258179   Test Acc:0.646
Epoch:3   Train Loss:1.5449037551879883   Test Acc:0.665
Epoch:4   Train Loss:1.3957326412200928   Test Acc:0.68
Epoch:5   Train Loss:1.2522907257080078   Test Acc:0.702
Epoch:6   Train Loss:1.1146392822265625   Test Acc:0.726
Epoch:7   Train Loss:0.9850963950157166   Test Acc:0.742
Epoch:8   Train Loss:0.865148663520813   Test Acc:0.753
Epoch:9   Train Loss:0.7546350359916687   Test Acc:0.766
Epoch:10   Train Loss:0.6532057523727417   Test Acc:0.779
Epoch:11   Train Loss:0.5612899661064148   Test Acc:0.789
Epoch:12   Train Loss:0.47923964262008667   Test Acc:0.798
Epoch:13   Train Loss:0.4067971706390381   Test Acc:0.802
Epoch:14   Train Loss:0.34367871284484863   Test Acc:0.802
Epoch:15   Train Loss:0.2893293499946594   Test Acc:0.801
Epoch:16   Train Loss:0.24299605190753937   Test Acc:0.805
Epoch:17   Train Loss:0.20382674038410187   Test Acc:0.806
Epoch:18   Train Loss:0.17094141244888306   Test Acc:0.804
Epoch:19   Train Loss:0.14352205395698547   Test Acc:0.803
Epoch:20   Train Loss:0.12082599848508835   Test Acc:0.803
Epoch:21   Train Loss:0.10212802141904831   Test Acc:0.805
Epoch:22   Train Loss:0.08679106831550598   Test Acc:0.809
Epoch:23   Train Loss:0.07417013496160507   Test Acc:0.81
Epoch:24   Train Loss:0.06373254209756851   Test Acc:0.808
Epoch:25   Train Loss:0.05506579577922821   Test Acc:0.804
Epoch:26   Train Loss:0.047877322882413864   Test Acc:0.802
Epoch:27   Train Loss:0.04192071035504341   Test Acc:0.802
Epoch:28   Train Loss:0.037012018263339996   Test Acc:0.8
Epoch:29   Train Loss:0.032964423298835754   Test Acc:0.798
Epoch:30   Train Loss:0.029622385278344154   Test Acc:0.798
Epoch:31   Train Loss:0.026850802823901176   Test Acc:0.799
Epoch:32   Train Loss:0.024533459916710854   Test Acc:0.798
Epoch:33   Train Loss:0.022576363757252693   Test Acc:0.798
Epoch:34   Train Loss:0.020912956446409225   Test Acc:0.798
Epoch:35   Train Loss:0.01949603296816349   Test Acc:0.797
Epoch:36   Train Loss:0.01829376071691513   Test Acc:0.796
Epoch:37   Train Loss:0.017278574407100677   Test Acc:0.795
Epoch:38   Train Loss:0.016427984461188316   Test Acc:0.794
Epoch:39   Train Loss:0.01572219282388687   Test Acc:0.794
Epoch:40   Train Loss:0.015136897563934326   Test Acc:0.794
Epoch:41   Train Loss:0.01465317141264677   Test Acc:0.794
Epoch:42   Train Loss:0.014255816116929054   Test Acc:0.794
Epoch:43   Train Loss:0.013931368477642536   Test Acc:0.796
Epoch:44   Train Loss:0.013672472909092903   Test Acc:0.795
Epoch:45   Train Loss:0.013474125415086746   Test Acc:0.796
Epoch:46   Train Loss:0.013331538066267967   Test Acc:0.796
Epoch:47   Train Loss:0.013239811174571514   Test Acc:0.795
Epoch:48   Train Loss:0.01319477055221796   Test Acc:0.797
Epoch:49   Train Loss:0.013191784732043743   Test Acc:0.797
Epoch:50   Train Loss:0.013221818953752518   Test Acc:0.797
Epoch:51   Train Loss:0.013279520906507969   Test Acc:0.8
Epoch:52   Train Loss:0.01336253248155117   Test Acc:0.8
Epoch:53   Train Loss:0.013466333039104939   Test Acc:0.801
Epoch:54   Train Loss:0.013588031753897667   Test Acc:0.803
Epoch:55   Train Loss:0.01372613850980997   Test Acc:0.802
Epoch:56   Train Loss:0.0138785345479846   Test Acc:0.802
Epoch:57   Train Loss:0.01404199842363596   Test Acc:0.803
Epoch:58   Train Loss:0.014217497780919075   Test Acc:0.803
Epoch:59   Train Loss:0.01439977902919054   Test Acc:0.804
Epoch:60   Train Loss:0.014586491510272026   Test Acc:0.803
Epoch:61   Train Loss:0.014775114133954048   Test Acc:0.803
Epoch:62   Train Loss:0.014964202418923378   Test Acc:0.803
Epoch:63   Train Loss:0.015152176842093468   Test Acc:0.804
Epoch:64   Train Loss:0.015336167998611927   Test Acc:0.805
Epoch:65   Train Loss:0.015514318831264973   Test Acc:0.805
Epoch:66   Train Loss:0.01568502001464367   Test Acc:0.805
Epoch:67   Train Loss:0.015847589820623398   Test Acc:0.804
Epoch:68   Train Loss:0.01600045897066593   Test Acc:0.804
Epoch:69   Train Loss:0.01614467054605484   Test Acc:0.805
Epoch:70   Train Loss:0.016276754438877106   Test Acc:0.805
Epoch:71   Train Loss:0.01639663800597191   Test Acc:0.805
Epoch:72   Train Loss:0.016504433006048203   Test Acc:0.805
Epoch:73   Train Loss:0.01659868098795414   Test Acc:0.805
Epoch:74   Train Loss:0.016679927706718445   Test Acc:0.803
Epoch:75   Train Loss:0.01674964651465416   Test Acc:0.803
Epoch:76   Train Loss:0.016807589679956436   Test Acc:0.803
Epoch:77   Train Loss:0.016852566972374916   Test Acc:0.802
Epoch:78   Train Loss:0.016885491088032722   Test Acc:0.802
Epoch:79   Train Loss:0.016908036544919014   Test Acc:0.802
Epoch:80   Train Loss:0.01692015491425991   Test Acc:0.8
Epoch:81   Train Loss:0.016921622678637505   Test Acc:0.8
Epoch:82   Train Loss:0.016912296414375305   Test Acc:0.8
Epoch:83   Train Loss:0.016894150525331497   Test Acc:0.8
Epoch:84   Train Loss:0.01686643622815609   Test Acc:0.8
Epoch:85   Train Loss:0.01682898961007595   Test Acc:0.8
Epoch:86   Train Loss:0.01678611896932125   Test Acc:0.799
Epoch:87   Train Loss:0.016737814992666245   Test Acc:0.799
Epoch:88   Train Loss:0.016682147979736328   Test Acc:0.799
Epoch:89   Train Loss:0.016619514673948288   Test Acc:0.799
Epoch:90   Train Loss:0.016552427783608437   Test Acc:0.8
Epoch:91   Train Loss:0.016480647027492523   Test Acc:0.8
Epoch:92   Train Loss:0.016405712813138962   Test Acc:0.8
Epoch:93   Train Loss:0.016329515725374222   Test Acc:0.8
Epoch:94   Train Loss:0.01624993048608303   Test Acc:0.8
Epoch:95   Train Loss:0.01616790145635605   Test Acc:0.8
Epoch:96   Train Loss:0.016083937138319016   Test Acc:0.799
Epoch:97   Train Loss:0.015997864305973053   Test Acc:0.801
Epoch:98   Train Loss:0.015910038724541664   Test Acc:0.801
Epoch:99   Train Loss:0.015820365399122238   Test Acc:0.801
Epoch:100   Train Loss:0.015730848535895348   Test Acc:0.801
Epoch:101   Train Loss:0.015641991049051285   Test Acc:0.801
Epoch:102   Train Loss:0.015551982447504997   Test Acc:0.801
Epoch:103   Train Loss:0.015460828319191933   Test Acc:0.801
Epoch:104   Train Loss:0.015370532870292664   Test Acc:0.801
Epoch:105   Train Loss:0.015280655585229397   Test Acc:0.801
Epoch:106   Train Loss:0.015191550366580486   Test Acc:0.801
Epoch:107   Train Loss:0.015103167854249477   Test Acc:0.801
Epoch:108   Train Loss:0.01501508243381977   Test Acc:0.801
Epoch:109   Train Loss:0.014928753487765789   Test Acc:0.801
Epoch:110   Train Loss:0.014843959361314774   Test Acc:0.801
Epoch:111   Train Loss:0.01476089283823967   Test Acc:0.801
Epoch:112   Train Loss:0.01467831153422594   Test Acc:0.801
Epoch:113   Train Loss:0.014596865512430668   Test Acc:0.801
Epoch:114   Train Loss:0.014517165720462799   Test Acc:0.801
Epoch:115   Train Loss:0.014438762329518795   Test Acc:0.801
Epoch:116   Train Loss:0.014360867440700531   Test Acc:0.802
Epoch:117   Train Loss:0.01428411528468132   Test Acc:0.803
Epoch:118   Train Loss:0.014208598993718624   Test Acc:0.802
Epoch:119   Train Loss:0.014135085977613926   Test Acc:0.802
Epoch:120   Train Loss:0.014063213020563126   Test Acc:0.802
Epoch:121   Train Loss:0.01399277988821268   Test Acc:0.802
Epoch:122   Train Loss:0.013923508115112782   Test Acc:0.802
Epoch:123   Train Loss:0.013854768127202988   Test Acc:0.802
Epoch:124   Train Loss:0.0137864351272583   Test Acc:0.803
Epoch:125   Train Loss:0.013720217160880566   Test Acc:0.803
Epoch:126   Train Loss:0.013654814101755619   Test Acc:0.804
Epoch:127   Train Loss:0.013590318150818348   Test Acc:0.804
Epoch:128   Train Loss:0.01352674700319767   Test Acc:0.804
Epoch:129   Train Loss:0.013463341630995274   Test Acc:0.804
Epoch:130   Train Loss:0.013401255011558533   Test Acc:0.804
Epoch:131   Train Loss:0.013341032899916172   Test Acc:0.805
Epoch:132   Train Loss:0.013280920684337616   Test Acc:0.806
Epoch:133   Train Loss:0.013221459463238716   Test Acc:0.806
Epoch:134   Train Loss:0.01316215842962265   Test Acc:0.806
Epoch:135   Train Loss:0.013103933073580265   Test Acc:0.807
Epoch:136   Train Loss:0.013046271167695522   Test Acc:0.807
Epoch:137   Train Loss:0.012989106588065624   Test Acc:0.807
Epoch:138   Train Loss:0.012933368794620037   Test Acc:0.807
Epoch:139   Train Loss:0.012878047302365303   Test Acc:0.806
Epoch:140   Train Loss:0.012822895310819149   Test Acc:0.806
Epoch:141   Train Loss:0.012768763117492199   Test Acc:0.807
Epoch:142   Train Loss:0.01271540205925703   Test Acc:0.807
Epoch:143   Train Loss:0.012662047520279884   Test Acc:0.807
Epoch:144   Train Loss:0.012610061094164848   Test Acc:0.807
Epoch:145   Train Loss:0.012558558024466038   Test Acc:0.808
Epoch:146   Train Loss:0.012508128769695759   Test Acc:0.808
Epoch:147   Train Loss:0.012457980774343014   Test Acc:0.808
Epoch:148   Train Loss:0.01240801066160202   Test Acc:0.808
Epoch:149   Train Loss:0.01235883217304945   Test Acc:0.808
Epoch:150   Train Loss:0.012310194782912731   Test Acc:0.808
Epoch:151   Train Loss:0.012262031435966492   Test Acc:0.808
Epoch:152   Train Loss:0.012214711867272854   Test Acc:0.808
Epoch:153   Train Loss:0.012168481945991516   Test Acc:0.808
Epoch:154   Train Loss:0.012121256440877914   Test Acc:0.808
Epoch:155   Train Loss:0.012074427679181099   Test Acc:0.809
Epoch:156   Train Loss:0.012028360739350319   Test Acc:0.809
Epoch:157   Train Loss:0.011983374133706093   Test Acc:0.809
Epoch:158   Train Loss:0.0119386101141572   Test Acc:0.809
Epoch:159   Train Loss:0.01189413946121931   Test Acc:0.809
Epoch:160   Train Loss:0.011850536800920963   Test Acc:0.809
Epoch:161   Train Loss:0.011807534843683243   Test Acc:0.809
Epoch:162   Train Loss:0.011764495633542538   Test Acc:0.809
Epoch:163   Train Loss:0.011721866205334663   Test Acc:0.809
Epoch:164   Train Loss:0.011679567396640778   Test Acc:0.809
Epoch:165   Train Loss:0.011638054624199867   Test Acc:0.809
Epoch:166   Train Loss:0.011597426608204842   Test Acc:0.809
Epoch:167   Train Loss:0.011557132937014103   Test Acc:0.809
Epoch:168   Train Loss:0.011517229489982128   Test Acc:0.809
Epoch:169   Train Loss:0.011477210558950901   Test Acc:0.809
Epoch:170   Train Loss:0.011438659392297268   Test Acc:0.809
Epoch:171   Train Loss:0.011400625109672546   Test Acc:0.809
Epoch:172   Train Loss:0.011362049728631973   Test Acc:0.807
Epoch:173   Train Loss:0.011323662474751472   Test Acc:0.807
Epoch:174   Train Loss:0.011285802349448204   Test Acc:0.807
Epoch:175   Train Loss:0.011248385533690453   Test Acc:0.807
Epoch:176   Train Loss:0.011211426928639412   Test Acc:0.807
Epoch:177   Train Loss:0.011174659244716167   Test Acc:0.807
Epoch:178   Train Loss:0.011138970032334328   Test Acc:0.809
Epoch:179   Train Loss:0.011103731580078602   Test Acc:0.808
Epoch:180   Train Loss:0.01106821745634079   Test Acc:0.808
Epoch:181   Train Loss:0.011033419519662857   Test Acc:0.808
Epoch:182   Train Loss:0.010998581536114216   Test Acc:0.808
Epoch:183   Train Loss:0.010963555425405502   Test Acc:0.808
Epoch:184   Train Loss:0.010929064825177193   Test Acc:0.809
Epoch:185   Train Loss:0.010895474813878536   Test Acc:0.809
Epoch:186   Train Loss:0.010861996561288834   Test Acc:0.809
Epoch:187   Train Loss:0.010829135775566101   Test Acc:0.809
Epoch:188   Train Loss:0.010796524584293365   Test Acc:0.809
Epoch:189   Train Loss:0.01076436322182417   Test Acc:0.809
Epoch:190   Train Loss:0.010732785798609257   Test Acc:0.809
Epoch:191   Train Loss:0.010701286606490612   Test Acc:0.809
Epoch:192   Train Loss:0.010670561343431473   Test Acc:0.81
Epoch:193   Train Loss:0.010640090331435204   Test Acc:0.81
Epoch:194   Train Loss:0.010609396733343601   Test Acc:0.81
Epoch:195   Train Loss:0.010578806512057781   Test Acc:0.81
Epoch:196   Train Loss:0.010548238642513752   Test Acc:0.81
Epoch:197   Train Loss:0.010518139228224754   Test Acc:0.81
Epoch:198   Train Loss:0.010489010252058506   Test Acc:0.81
Epoch:199   Train Loss:0.010459662415087223   Test Acc:0.81进程已结束，退出代码为 0
http://www.xdnf.cn/news/18094.html

相关文章：

  • 摩搭api 实现
  • 【最后203篇系列】032 OpenAI格式调用多模型实验
  • 博弈论07——Lemke-Howson 算法
  • Redis入门与背景详解：构建高并发、高可用系统的关键基石
  • SpringBoot自动配置原理（二）
  • [Linux] RAID存储技术
  • TDengine 3.3.7.0 版新功能（BLOB 数据类型）
  • LangGraph实战指南：如何构建一个深度研究Agent
  • 测试工程师的AI转型指南：从工具使用到测试策略重构
  • MySQL 配置性能优化赛技术指南
  • H20芯片与中国的科技自立：一场隐形的博弈
  • K8S HPA 弹性水平扩缩容 Pod 详解
  • 力扣 hot100 Day75
  • C#WPF实战出真汁08--【消费开单】--餐桌面板展示
  • implement copy file content to clipboard on Windows
  • 某智慧教育平台登录流程分析
  • php版的FormCreate使用注意事项
  • 树、哈夫曼树以及二叉树的各种操作
  • 如何使用嵌入模型创建本地知识库Demo
  • Rocky Linux 9.2：从 /home 分区释放 10G 空间扩容到 / 根分区
  • 二进制为什么使用记事本读取会出乱码
  • eChart饼环pie中间显示总数_2个以上0值不挤掉
  • 0815 UDP通信协议TCP并发服务器
  • Mac （三）如何设置环境变量
  • 深入理解 Python 元类中的 __prepare__ 方法：掌控类属性定义顺序的艺术
  • docker镜像解决的一些问题
  • 双重调度（Double Dispatch）：《More Effective C++》条款31
  • [Linux] Linux网络管理
  • 16-集合的Stream编程
  • 宋红康 JVM 笔记 Day03｜内存结构概述、类加载器与类的加载过程、类加载器分类