A Simple Training Strategy for Graph Autoencoder

被引：10

作者：

Wang, Yingfeng ^{[1
]}

Xu, Biyun ^{[2
]}

Kwak, Myungjae ^{[1
]}

Zeng, Xiaoqin ^{[3
]}

机构：

[1] Middle Georgia State Univ, 100 Univ Pkwy, Macon, GA 31206 USA

[2] Beijing Kubao Technol Co, 1089 Huihe South St, Beijing, Peoples R China

[3] Hohai Univ, 1 Xikang Rd, Nanjing, Jiangsu, Peoples R China

来源：

ICMLC 2020: 2020 12TH INTERNATIONAL CONFERENCE ON MACHINE LEARNING AND COMPUTING | 2018年

基金：

美国国家科学基金会;

关键词：

Autoencoder; Graph Convolutional Network; Perturbation; Training Algorithm;

D O I：

10.1145/3383972.3383985

中图分类号：

TP301 [理论、方法];

学科分类号：

081202 ;

摘要：

Graph autoencoder can map graph data into a low-dimensional space. It is a powerful graph embedding method applied in graph analytics to reduce the computational cost. The training algorithm of a graph autoencoder searches the weight setting for preserving most graph information of the graph data with reduced dimensionality. This paper presents a simple training strategy, which can improve the training performance without significantly increasing time complexity. This strategy can flexibly fit many existing training algorithms. The experimental results confirm the effectiveness of this strategy.

引用

页码：341 / 345

页数：5

共 25 条

[11]

Kingma DP, 2014, Adv Neural Inf Process Syst, P14

[12]

Kipf TN., 2016, NEURIPS WORKSH BAYES

[13]

Li Y, 2016, P 4 INT C LEARN REPR

[14] ANALYZING A PORTION OF THE ROC CURVE [J].

MCCLISH, DK .

MEDICAL DECISION MAKING, 1989, 9 (03) :190-195

[15]

McDowell LK, 2009, J MACH LEARN RES, V10, P2777

[16]

Pan SR, 2018, PROCEEDINGS OF THE TWENTY-SEVENTH INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE, P2609

[17] The Graph Neural Network Model [J].

Scarselli, Franco ;

Gori, Marco ;

Tsoi, Ah Chung ;

Hagenbuchner, Markus ;

Monfardini, Gabriele .

IEEE TRANSACTIONS ON NEURAL NETWORKS, 2009, 20 (01) :61-80

[18]

Thomas M.W., 2017, 5 INT C LEARN REPR I, P1, DOI DOI 10.48550/ARXIV.1609.02907

[19] Deep Recursive Network Embedding with Regular Equivalence [J].

Tu, Ke ;

Cui, Peng ;

Wang, Xiao ;

Yu, Philip S. ;

Zhu, Wenwu .

KDD'18: PROCEEDINGS OF THE 24TH ACM SIGKDD INTERNATIONAL CONFERENCE ON KNOWLEDGE DISCOVERY & DATA MINING, 2018, :2357-2366

[20] MGAE: Marginalized Graph Autoencoder for Graph Clustering [J].

Wang, Chun ;

Pan, Shirui ;

Long, Guodong ;

Zhu, Xingquan ;

Jiang, Jing .

CIKM'17: PROCEEDINGS OF THE 2017 ACM CONFERENCE ON INFORMATION AND KNOWLEDGE MANAGEMENT, 2017, :889-898

← 1 2 3 →