存档0427

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
【项目介绍】 基于Keras+cora和citeseer数据集实现GAT训练及节点分类测试python源码+数据集+项目说明.zip 环境: CUDA:11.6.134 cuDNN:8.4.0 keras:2.9.0 tensorflow:2.9.1 注意: 项目内目录中两个文件夹: 1. /datasets:将数据集文件解压至此 2. /save_models:保存训练好的模型权重文件,包括生成器权重和判别器权重两个文件 GAT概述 图神经网络(Graph Neural Network, GNN)是指神经网络在图上应用的模型的统称,根据采用的技术不同和分类方法的不同, 又可以分为下图中的不同种类,例如从传播的方式来看,图神经网络可以分为图卷积神经网络(GCN),图注意力网络(GAT),Graph LSTM等等 图注意力网络(Graph Attention Network, GAT),一种基于图结构数据的新型神经网络架构,利用隐藏的自我注意层来解决之前基于图卷积或其近似的方法的不足。通过堆叠层,节点能够参与到邻居的特征,可以(隐式地)为邻域中的不同节点指定不同的权值,而不需要任何代价高昂的矩阵操作(如反转),也不需要预先知道图的结构。通过这种方法,该模型克服了基于频谱的故神经网络的几个关键挑战,并使得模型适用于归纳和推理问题。 数据集: cora:包含2708篇科学出版物网络,共有5429条边,总共7种类别。 数据集中的每个出版物都由一个 0/1 值的词向量描述,表示字典中相应词的缺失/存在。 该词典由 1433 个独特的词组成。 链接:https://pan.baidu.com/s/1u7v3oJcTvnFWAhHdSLHwtA?pwd=52dl 提取码:52dl citeseer:包含3312个节点,4723条边构成的引文网络。标签共6个类别。数据集的特征维度是3703维。 链接:https://pan.baidu.com/s/11n2AQCVSV6OevSkUhYWcNg?pwd=52dl 提取码:52dl 通过测试,采用以下设置: train_nodes = 140# 训练节点数量 epochs = 100# 迭代次数 hidden_dim=128# 隐层维度 att_heads=6# 注意力头数量 dropout_rate = 0.5# dropout概率率 Adam LR = 5e-3# 学习率 GAT在cora数据集和citeseer数据集上具有70%和80%左右的准确率,上面参数随便设置的,调好超参数应该还能提高一点。 【备注】 1.项目代码均经过功能验证,确保稳定可靠运行。欢迎下载食用体验! 2.主要针对各个计算机相关专业,包括计算机科学、信息安全、数据科学与大数据技术、人工智能、通信、物联网等领域的在校学生、专业教师、企业员工。 3.项目具有丰富的拓展空间,不仅可作为入门进阶,也可作为毕设、课程设计、大作业、初期项目立项演示等用途。 4.鼓励大家基于此进行二次开发。在使用过程中,如有问题或建议,请及时沟通。 5.期待你能在项目中找到乐趣和灵感,也欢迎你的分享和反馈!
bhaaAThangemMkaz2tQhq3/f7dZ7 Vj29WeGHjuupj/AhYqymjAuokhYi X1T/fG+q1yR22PdcEP39dxU3ovEo lLLzwIZlQr9oJYwUf5eG4x5e1bMP nfIaIl8reszQPHFNbrxBjCVrBEXL TyBLehxzUpVI672t3LjX8q4ytCRC ezcoZQfMAc0DulJzNREbwlcf0Rco N0zW3ojBUxGsplBBZZPrD7/bPhkM rjtS8Bxym+HeV1TJDTXuiL8QLg== ====================================== fSbC9xxy7ga57DOrLz9mMqYSFx8S OIWiOBI61uJ562GINVTVNBDaGBIl Cuni0a1iUAF88I6zrHzH9Wzcv4h+ h3wTGgOX8mYgN0jEEOmzI1DLx0DJ 8t94hUJ5O0I99Ts6iqvHhw24jYfI h++JG/BzVLnaSDbNdqH9iYbXZ2Pc eFZAZe6hzPQ9DREtkEvg7rAul/Fw ZptCP1WFfKkfgviwY8i67NucJe2L UTZFKoX5MGSuP9GGJnouQg/P1SII 3PCKXGoUv0hBoTiMLXBlJWpaNg== ============================================= 0W01VaFQPp/XRYgXjOFSUY7hZ5Cc W9f35KH4bym2jmhYygsHyNsr47Hs 7OBjd7QH+ALAKqZdDAh/AfG3V/8Q his64uZEhglY8B8yu5PvN49tOO05 XG7Qyz3skOK3An9Zoc/xSA1fLGeq PioqBj/khHWF1LK0lBi00LOLFXzs oQtoMi1Ol+Ehxz8URH3kQqN09raw iM7DrTxZe1oVoOjTPDerwPOLULI1 6EnbdBu92YSk0tYQw0BmVss2G6ak GjaDrPWrDWP3uip0OOX4UXNggQ== =============================================== H4nLHG3xsjy5l4QMEsSpaYAyggPp 9OdFDMG0Y1XM7VqTyg+GAAm8V/eP 9zuAwQ5/9tA8GQeOlavXxIDkMsY7 I4UAy4LmhUU2vLtEkgjtMhSd3nFT OGe2bOmGvoMCPts/1eQx5YFMet9K AjkZPI1BR8/fm7Mb0bQOPeJQyilC T1/4k4MfqnQkub06+ADZ7pg3WdkC jneqAVvKPaM2Kf55cGDsLmo3Xxck ZHpzPKJDISjEQ+qFadS8Z9Ao0HHK 31JWyvw0Qr/9d0Ku10+W6DrhZw== ================================================= DZaNXFw2rfRrF3xlukPGGiilkurQ r7vKuSPCUEdAoRWRJmjnADY0Ink3 Nr9qD3CM5+v9PYJ4uf/b9jShtlYH KV16XPA+On7CjH8zqxywbz2ci85i f0va/I0DOvGhpsww1jAJCJWC4A+z ZjTMzA0+SEQcKyHzOUe89eA+c/Gq FZBmitn5u602wlMj2KtvMSN8EgSc wDVxI+FSi0U/b89fRaUMJRbUadm0 uZHKMPdlaykMQPwZFN8G561OFCU/ rbFQlatd0JFPnLKxjNsjva04LQ== =============================================== hFeWdc8oboH/QwzaaLhrH6CgRo6g nBXqa0bjHJdEYTIzMdqnkEBvygf+ xJeO2zRqajhnH+S1wZWeNUwXLwZm KvS8LCvwCf1WBqzsRMuDOTQnac5d RGf80G986szBG8kzkhKjC4Bx3uLy NomiYSHrGBoDacd5wPCqFMoTsaMC CjVMabK/JyQyAeuULOf8lW0WiXTo TKH7k1yGW6plaQKWGIqBZuYlYk5H tFm8hdKXaxbeiiREFl7ni1G9lP7j 1Gdf/cmb+fq2sKrSgZyLEjQLFA== =============================================== Y2CjWsyI23Fa0G1pban8ep83zQAI 7mERztK7a3ZufX0LWeAe/y50DURN kOOlakZ0dgkoipf/6KsNEEm2aptm L3/dXG1jKphojzSwNsS4PHY4qGKA 9vHBEAsDLlwGrLFfmdtKOCrgNG6V ab7clt+yRFuxctwefegNu02YdlGC wEKI2MP9QcMXl/aFLkFsz3mBFn1t YZtjYFswD+7zJ6KETy1cRK0TG/UK /6GoBQtjLIeCjNBbrEWvOm6XunsO jCaOhBeccOe3lIUCbJEwSFKAGA== ============================================= 3AFJhTfHVl0MBUFC1NEgO34cyEqR saJhnLE8X1K0OU5wU630QIlLyAiu pMGYcSoTEanqEaBvKiXFfMvqbTM+ X87dmWn6bicol0G9APDw1GKO4A4g Zyc8z47uRSNgqD9moJGFIdZZ+l1b +/PCg5MKBBhDhc6VDBKJ3hdVTu6L sW3+fY+9awZvyMMrXNu3STU0z0TX kHVOuMI+XqKeywPGEClGO/U5vgN7 DbL2xQm0Czm89V3fPnY+03gVOkP2 dvBKP4EqwgCXuxJVDEbC8T0bPQ== ============================================= SoydMqMl+/Gpf41SeCYSamtq3urK wqREsSLloXTdR1DoEd6sFEtxWH+0 w/nC12MfI+TzlhOuvit3bqOzUSrq XTPipT1EA1YfelWllo3inkHWZ8Bd IG4b/GCULr4ufAKEtZnyXVHeUPoo /HCkWPViDUwQUbQqZhcu7baPDnAo fGQlW5++dTXq3yIZTxNs8tpffgEk 5oemtUyJhXaAEas6KQKw/vEFLPho 48B3tuL31SNvluTSp3O3y8qZvSEV h3GE09Cp7E4SWKkv6rB9VG8Vcg== =============================================== Xzlv+2HZXpuIUhAUGVYmjhw6ncp9 b7qlOIwQa4BiPlhJl61ma2fUW1s4 BvNSzx0WgiHGB86X6VZ1ix3bXZop eq4MxQFHKtaLj60Y6jduOX8Id9Is oBIKLLkuRijSbZJpBxIqHLrz+2C/ nym1n8XbWYagYJ1z4OqNxjkrCGfW LoSPC9jAtivGSk3T5lDsUzh6MEqF bSd6Bjce5tqDgX9AV0DM3icFg3dF /u9XXp+72DOCT9FStV01UwID6L/O zzokhH37OQL8irSbjE1FdGW+TQ== ========================================== BJNZHBPeVX4GKH3AbulBcezga945 qNEKaHM8PpixioQLO0rZ57tUUQ7l jkWF/5Vh6pWVSJTmNhM54CkFDq3a M6EVzfPJT00+YS66mmQe4FI85zaj cvyNyiaKqBh8XjwxXhFs6oHjYJJ4 MAWRTTvOVXDH29E7J30UQkz4qFWD 0445Aj0clJCfu4WAmGq7TmAWF8H2 g+k5VUTx8aBnRS/WVMG3iyryv+3h W2S4LjfsnF1Zsaf4D4N9p5Uqr2rI Lvhaj80tbxjTja+MLVhNr4GURQ== ================================================= Lv/+h44wqgNXFicAJPjLZFqwpCoF mnn2JvfPTtQ9kEX51XdLNImcH6Pe bKp1KUMMGaUWSA5/NRMqqmoVaJQm Fh3B9G2jCyVfYeag+EFRCI/ByCpx Jg+yvnE4i9NUS1mKuBXD/QR5uEBk OsiAzv8DgQORN/sbIkjUKX7vwvzG 6QEaDIPaWzKqPV7RIctjkUt4+5Av r5oBre/6Mg1Q1GIhBftZkD1UtPm3 /3oDrT3o+2SlPNwElOcmypfkMHkk jAVK5+vfb8lOPKou3lmzUPZPKw== ============================================== zPH5/yKwB7akS0BMDoMsIRoIA0dN rE5EB8qSLhyCiwp+oZVZOYm7fH/o z7odFfCIGuTPHwUOB7r2u6AT1v0a i0vSNJXevY/dv/EY2hCClIhxTohs GUE0VLLV3vZJ0+AalJPuXfP3v513 PQgPG0jLLMHoYm2gC0eZU4TCQx0B r4JJo5Ne0CH9wGfwiPdkIul0NhHU uQGP78v0FIsTBbQvDuing6q3msxX GQ+GeVUmshiefNvZ8MzB86ug0PaL hgtayc9N3lwGEf23C2S/MZW6bQ== ================================================== pF+PyM1Vz5eTwApjWBBOC8UtZmOV 0xsmgjgvdTLVMXc8pLbhdsDkvMyI sgBv8WZuftr4FdKRxQVXVr+j5MAi ai3I9jJ4msA80i5ITzj+UykFuXeJ M8AB68iTiN80FjoKA1o2aJcjzc1J ffxHJCavDvlFG7KnfbuM9VjLHIbQ 9lkXTIQX20o6qYOPmPpOG7JqeVLe H4XZLbTjzJp81h5YY9dVISCKbf01 1Wl6vlh3L1Q1IFSnIfaQpTZwixHw bJZE8tuy2STl7G17uiqjuvcmdg==
评论 1
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值