DuaLGR: Dual Label-Guided Graph Refinement for Multi-View Graph Clustering

This is the source code for paper: Dual Label-Guided Graph Refinement for Multi-View Graph Clustering, accepted at AAAI 2023.

Requirements

• 'requirements.txt'
• The experiments are conducted on a Linux machine with a NVIDIA GeForce RTX 3070 GPU and Intel(R) Xeon(R) E5-2678 v3 @ 2.50GHz CPU.

Datasets

ACM, ACM (HR 0.00) and ACM (HR 0.20) datasets are included in ./data/, Texas, DBLP and Chameleon are publicly available, and other synthetic data will be published in the future.

Raw data

Dataset	#Clusters	#Nodes	#Features	Graphs	HR
ACM	3	3025	1830	$\mathcal{G}^1$ $\mathcal{G}^2$	0.82 0.64
DBLP	4	4057	334	$\mathcal{G}^1$ $\mathcal{G}^2$ $\mathcal{G}^3$	0.80 0.67 0.32
Texas	5	183	1703	$\mathcal{G}^1$	0.09
Chameleon	5	22777	2325	$\mathcal{G}^1$	0.23

Synthetic data

Dataset	#Clusters	#Nodes	#Features	Graphs	HR
ACM (HR 0.00)	3	3025	1830	$\mathcal{G}^1$ $\mathcal{G}^2$	0.00 0.00
ACM (HR 0.10)	3	3025	1830	$\mathcal{G}^1$ $\mathcal{G}^2$	0.10 0.10
ACM (HR 0.20)	3	3025	1830	$\mathcal{G}^1$ $\mathcal{G}^2$	0.20 0.20
ACM (HR 0.30)	3	3025	1830	$\mathcal{G}^1$ $\mathcal{G}^2$	0.30 0.30
ACM (HR 0.40)	3	3025	1830	$\mathcal{G}^1$ $\mathcal{G}^2$	0.40 0.40
ACM (HR 0.50)	3	3025	1830	$\mathcal{G}^1$ $\mathcal{G}^2$	0.50 0.50

Test DuaLGR

# Test DuaLGR on ACM dataset
python DuaLGR.py --dataset 'acm' --train False --use_cuda True --cuda_device 0

# Test DuaLGR on DBLP dataset
python DuaLGR.py --dataset 'dblp' --train False --use_cuda True --cuda_device 0

# Test DuaLGR on Texas dataset
python DuaLGR.py --dataset 'texas' --train False --use_cuda True --cuda_device 0

# Test DuaLGR on Chameleon dataset
python DuaLGR.py --dataset 'chameleon' --train False --use_cuda True --cuda_device 0

# Test DuaLGR on synthetic ACM dataset with HR 0.00
python DuaLGR.py --dataset 'acm00' --train False --use_cuda True --cuda_device 0

# Test DuaLGR on synthetic ACM dataset with HR 0.10
python DuaLGR.py --dataset 'acm01' --train False --use_cuda True --cuda_device 0

# Test DuaLGR on synthetic ACM dataset with HR 0.20
python DuaLGR.py --dataset 'acm02' --train False --use_cuda True --cuda_device 0

# Test DuaLGR on synthetic ACM dataset with HR 0.30
python DuaLGR.py --dataset 'acm03' --train False --use_cuda True --cuda_device 0

# Test DuaLGR on synthetic ACM dataset with HR 0.40
python DuaLGR.py --dataset 'acm04' --train False --use_cuda True --cuda_device 0

# Test DuaLGR on synthetic ACM dataset with HR 0.50
python DuaLGR.py --dataset 'acm05' --train False --use_cuda True --cuda_device 0

Train DuaLGR

# Train DuaLGR on ACM dataset
python DuaLGR.py --dataset 'acm' --train True --use_cuda True --cuda_device 0

# Train DuaLGR on DBLP dataset
python DuaLGR.py --dataset 'dblp' --train True --use_cuda True --cuda_device 0

# Train DuaLGR on Texas dataset
python DuaLGR.py --dataset 'texas' --train True --use_cuda True --cuda_device 0

# Train DuaLGR on Chameleon dataset
python DuaLGR.py --dataset 'chameleon' --train True --use_cuda True --cuda_device 0

# Train DuaLGR on synthetic ACM dataset with HR 0.00
python DuaLGR.py --dataset 'acm00' --train True --use_cuda True --cuda_device 0

# Train DuaLGR on synthetic ACM dataset with HR 0.10
python DuaLGR.py --dataset 'acm01' --train True --use_cuda True --cuda_device 0

# Train DuaLGR on synthetic ACM dataset with HR 0.20
python DuaLGR.py --dataset 'acm02' --train True --use_cuda True --cuda_device 0

# Train DuaLGR on synthetic ACM dataset with HR 0.30
python DuaLGR.py --dataset 'acm03' --train True --use_cuda True --cuda_device 0

# Train DuaLGR on synthetic ACM dataset with HR 0.40
python DuaLGR.py --dataset 'acm04' --train True --use_cuda True --cuda_device 0

# Train DuaLGR on synthetic ACM dataset with HR 0.50
python DuaLGR.py --dataset 'acm05' --train True --use_cuda True --cuda_device 0

Parameters: More parameters and descriptions can be found in the script and paper.

Results of DuaLGR

	NMI%	ARI%	ACC%	F1%
ACM	73.2	79.4	92.7	92.7
DBLP	75.5	81.7	92.4	91.8
Texas	36.6	27.8	57.4	43.3
Chameleon	18.6	13.5	42.1	41.1
ACM (HR 0.00)	61.6	68.5	88.3	88.3
ACM (HR 0.10)	62.0	69.1	88.6	88.5
ACM (HR 0.20)	62.9	70.0	89.0	89.0
ACM (HR 0.30)	63.7	71.0	89.4	89.4
ACM (HR 0.40)	93.7	96.5	98.8	98.8
ACM (HR 0.50)	99.5	99.8	99.9	99.9