H. Kashima, K. Tsuda, and A. Inokuchi. Proceedings of the Twentieth International Conference on Machine Learning, page 321--328. AAAI Press, (2003)
Abstract
A new kernel function between two labeled graphs is presented. Feature vectors are defined as the counts of label paths produced by random walks on graphs. The kernel computation finally boils down to obtaining the stationary state of a discrete-time linear system, thus is efficiently performed by solving simultaneous linear equations. Our kernel is based on an infinite dimensional feature space, so it is fundamentally different from other string or tree kernels based on dynamic programming. We will present promising empirical results in classification of chemical compounds. 1 1.
Description
CiteSeerX — Marginalized kernels between labeled graphs
%0 Conference Paper
%1 Kashima03marginalizedkernels
%A Kashima, Hisashi
%A Tsuda, Koji
%A Inokuchi, Akihiro
%B Proceedings of the Twentieth International Conference on Machine Learning
%D 2003
%I AAAI Press
%K graph-kernel
%P 321--328
%T Marginalized kernels between labeled graphs
%U http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.90.7556
%X A new kernel function between two labeled graphs is presented. Feature vectors are defined as the counts of label paths produced by random walks on graphs. The kernel computation finally boils down to obtaining the stationary state of a discrete-time linear system, thus is efficiently performed by solving simultaneous linear equations. Our kernel is based on an infinite dimensional feature space, so it is fundamentally different from other string or tree kernels based on dynamic programming. We will present promising empirical results in classification of chemical compounds. 1 1.
@inproceedings{Kashima03marginalizedkernels,
abstract = {A new kernel function between two labeled graphs is presented. Feature vectors are defined as the counts of label paths produced by random walks on graphs. The kernel computation finally boils down to obtaining the stationary state of a discrete-time linear system, thus is efficiently performed by solving simultaneous linear equations. Our kernel is based on an infinite dimensional feature space, so it is fundamentally different from other string or tree kernels based on dynamic programming. We will present promising empirical results in classification of chemical compounds. 1 1.},
added-at = {2010-11-24T11:47:53.000+0100},
author = {Kashima, Hisashi and Tsuda, Koji and Inokuchi, Akihiro},
biburl = {https://www.bibsonomy.org/bibtex/2a574cec771c2c58f24496036219d7ca9/utahell},
booktitle = {Proceedings of the Twentieth International Conference on Machine Learning},
description = {CiteSeerX — Marginalized kernels between labeled graphs},
interhash = {a899bc24b3cd3044df5223f39f871761},
intrahash = {a574cec771c2c58f24496036219d7ca9},
keywords = {graph-kernel},
pages = {321--328},
publisher = {AAAI Press},
timestamp = {2010-11-24T11:47:53.000+0100},
title = {Marginalized kernels between labeled graphs},
url = {http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.90.7556},
year = 2003
}