/Alternating_Diffusion_Process

A demo for the alternating diffusion process

Primary LanguageMATLABMIT LicenseMIT

Alternating_Diffusion_Process

By Qilin Li, Senjian An, Wanquan Liu, Ling Li

Curtin University.

Introduction

This is an matlab implementation of the Alternating_Diffusion_Process (ADP). ADP is essentially an affinity learning method embeded in the graph-based semi-supervised learning (GSSL). ADP integrates the affinity learning and label propagation of GSSL into a feedback framework, in which the predicted pseudo labels can be used to supervise the affinity learning process.

Usage

There are two demo test cases, simply run

testCase_COIL20.m
testCase_synthetic_fiveCircles.m

Reference

@article{li2019semi,
	title={Semi-supervised Learning on Graph with an Alternating Diffusion Process},
	author={Li, Qilin and An, Senjian and Li, Ling and Liu, Wanquan},
	journal={arXiv preprint arXiv:1902.06105},
	year={2019}
}