摘要
现有网络表示学习算法主要针对网络结构信息进行表示学习,而忽略现实网络中丰富的节点文本属性信息。为有效融合网络结构信息和节点文本属性信息进行表示学习,提出一种新的网络表示学习算法。为实现两方面信息在训练过程中的相互约束,建立基于参数共享的共耦神经网络训练模型,并利用负采样和随机梯度下降的优化策略实现训练过程的快速收敛。实验结果表明,与Doc2Vec算法、DeepWalk算法、DW+D2V算法和TADW算法相比,该算法的分类性能更好。
The existing network representation learning algorithms mainly focus on how to represent the network structure information,and ignore the abundant textual attribute information of nodes in real network.In order to incorporate network structure information and nodes’textual attribute information,this paper presents a novel network representation learning algorithm incorporating with nodes’textual attribute information.As to achieve mutual restraint of the two part of network information during the training process,this algorithm constructs a coupled neural network training model based on parameter sharing stratagem.It applies optimization strategy based on negative sample and stochastic gradient descent to achieve rapid convergence of the training process,and performs an experimental evaluation of node classification.Experimental results demonstrate that compared with Doc2Vec algorithm,DeepWalk algorithm,DW+D2V algorithm and TADW algorithm,the classification performance of the proposed algorithm is better.
作者
刘正铭
马宏
刘树新
杨奕卓
李星
LIU Zhengming;MA Hong;LIU Shuxin;YANG Yizhuo;LI Xing(National Digital Switching System Engineering and Technological R&D Center,Zhengzhou 450002,China)
出处
《计算机工程》
CAS
CSCD
北大核心
2018年第11期165-171,共7页
Computer Engineering
基金
国家自然科学基金(61521003)
关键词
复杂网络
网络表示学习
信息融合
文本属性信息
神经网络
complex network
network representation learning
information fusion
textual attribute information
neural network
