Deep Renyi entropy graph kernel

doi:10.1016/j.patcog.2020.107668

Deep Renyi entropy graph kernel

2021-03-01

23

PRE

AI

DOI：10.1016/j.patcog.2020.107668

Original

Original request for help

Save

Abstract

En 中文

Graph kernels are applied heavily for the classification of structured data. In this paper, we propose a deep Renyi entropy graph kernel for this purpose. We gauge the deep information through a family of h-layer expansion subgraphs rooted at a vertex, and define a h-layer depth-based second-order Renyi entropy representation for each vertex. The second-order Renyi entropy representation is used together with Euclidean distance to build a deep second-order Renyi entropy graph kernel (SREGK). For graphs with n vertices, the time complexity for our kernel is O(n(3)). This low-order polynomial complexity enables our subgraph kernels to easily scale up to graphs of reasonably large sizes and thus overcome the size limits arising in state-of-the-art graph kernels. Experimental results on fourteen real world graph datasets are shown to demonstrate the overall superior performance of our approach over a number of state-of-the-art methods. (C) 2020 Elsevier Ltd. All rights reserved.

Keywords:

Shannon entropy

Renyi entropy

Deep representation

Graph kernel

Graph classification

Deep Renyi entropy graph kernel

Abstract

AI Summary

Journal

Organization