This repository contains MATLAB code for 7 multi-view spectral clustering algorithms (and a single-view spectral clustering algorithm) used for comparison in our paper "Multi-view Graph Learning by Joint Modeling of Consistency and Inconsistency", which is the follow-up work of our ICDM paper "Consistency Meets Inconsistency: A Unified Graph Learning Framework for Multi-view Clustering".
The code of some algorithms was gathered from the websites of the authors of the original papers and was later fixed and optimized by us.
Please see our paper for the details of these algorithms (the folder names correspond to the abbreviations for the algorithms in our paper, namely, AASC, AWP, CoReg, MCGC, MVGL, RMSC, and WMSC). In each of these folders, there is a main file xxx_main.m
for the algorithm where xxx
is the algorithm name.
The original papers for the 7 multi-view spectral clustering algorithms and the single-view spectral clustering (SC) algorithm are:
For the code of our Multi-view Graph Learning algorithm, please see this repository.
All datasets used in our paper are available at Baidu Cloud with code pqti
and Google Drive. Each dataset is a mat file containing 2 variables fea
(i.e., a MATLAB cell of features) and gt
(i.e., ground truth label), except the file flower17.mat
which contains a cell of distance matrices and ground truth since features are unavailable.
flower17.mat
should be squared before passing them into the SGF and DGF functions, and the string original
should be passed into the functions as the metric parameter.Reuters
, Reuters-21578
, BBCSport
, and CiteSeer
are text datasets with word frequence as features and thus should be used with the cosine
metric for computing distance matrices.addpath('MinMaxSelection'); addpath('utils')
command in MATLAB command window.addpath('MinMaxSelection'); addpath('utils')
command in MATLAB command window. Then recompile the helper functions by running minmax_install
.The file test.m
contains examples to use all the algorithms.
此处可能存在不合适展示的内容,页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容,可点击提交进行申诉,我们将尽快为您处理。