Nonparametric small random networks for graph-structured pattern recognition

被引：6

作者：

Trentin, Edmondo ^{[1
]}

Di Iorio, Ernesto ^{[2
]}

机构：

[1] DIISM Univ Siena, Via Roma 56, I-53100 Siena, Italy

[2] QuestIT Srl, Via Leonida Cialfi 23, I-53100 Siena, Italy

来源：

NEUROCOMPUTING | 2018年 / 313卷

关键词：

Graph classification; Structured pattern recognition; Graph neural network; Random network; Relational learning; NEURAL-NETWORKS; CLASSIFICATION; MODEL;

D O I：

10.1016/j.neucom.2018.05.095

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

Taking inspiration from the probabilistic principles underlying the topological regularities observed in random networks, the paper presents a simple and efficient Bayesian framework for the classification of (small) labeled random networks. The proposed "graphical model" relies on a Parzen window estimate of the pairwise vertex-vertex probability distribution under an implicit Markov assumption. Experiments show that, in spite of its simplicity, the approach is at least as accurate as the state-of-the-art machines. The highest average recognition accuracies to date were obtained on the friendly + unfriendly Mutagenesis classification task. (C) 2018 Elsevier B.V. All rights reserved.

引用

页码：14 / 24

页数：11