Skip to content
This repository has been archived by the owner on Feb 25, 2022. It is now read-only.

hazdzz/MGCs

Repository files navigation

Magnetic Graph Convolutional Networks v1

Magnetic Eigenmaps

A directed 4-cycle

About

The official PyTorch implementation for the paper sMGC: A Complex-Valued Graph Convolutional Network via Magnetic Laplacian for Directed Graphs.

Requirements

To install requirements:

pip3 install -r requirements.txt

Results

Node classification accuracy in Citation networks (%)

Model CoRA CiteSeer PubMed
GAT 82.60 ± 0.40 70.45 ± 0.25 77.45 ± 0.45
sMGC 82.70 ± 0.00 73.30 ± 0.00 79.90 ± 0.10
MGC 82.50 ± 1.00 71.25 ± 0.95 79.70 ± 0.40

Node classification accuracy in WebKB (%)

Model Cornell Texas Washington Wisconsin
GAT 41.03 ± 0.00 52.63 ± 2.63 63.04 ± 0.00 56.61 ± 1.88
sMGC 73.08 ± 1.28 71.05 ± 0.00 68.48 ± 3.26 80.19 ± 2.83
MGC 80.77 ± 3.85 82.90 ± 1.31 70.66 ± 1.08 87.74 ± 2.83

Reproduce experiment results

sMGC

CoRA:

python3 main_smgc.py --mode='test' --seed=100 --dataset_config_path='./config/data/cora.ini' --alpha=0.03 --t=8.05 --K=38

CiteSeer:

python3 main_smgc.py --mode='test' --seed=100 --dataset_config_path='./config/data/citeseer.ini' --alpha=0.01 --t=5.16 --K=40

PubMed:

python3 main_smgc.py --mode='test' --seed=100 --dataset_config_path='./config/data/pubmed.ini' --alpha=0.01 --t=5.95 --K=25

Cornell:

python3 main_smgc.py --mode='test' --seed=100 --dataset_config_path='./config/data/cornell.ini' --alpha=0.95 --t=45.32 --K=12

Texas:

python3 main_smgc.py --mode='test' --seed=100 --dataset_config_path='./config/data/texas.ini' --alpha=0.71 --t=45.08 --K=23

Washington:

python3 main_smgc.py --mode='test' --seed=100 --dataset_config_path='./config/data/washington.ini' --alpha=0.77 --t=45.95 --K=44

Wisconsin:

python3 main_smgc.py --mode='test' --seed=100 --dataset_config_path='./config/data/wisconsin.ini' --alpha=0.93 --t=25.76 --K=34

MGC

CoRA:

python3 main_mgc.py --mode='test' --seed=100 --dataset_config_path='./config/data/cora.ini' --alpha=0.08 --t=5.85 --K=10 --droprate=0.4

CiteSeer:

python3 main_mgc.py --mode='test' --seed=100 --dataset_config_path='./config/data/citeseer.ini' --alpha=0.01 --t=25.95 --K=35 --droprate=0.3

PubMed:

python3 main_mgc.py --mode='test' --seed=100 --dataset_config_path='./config/data/pubmed.ini' --alpha=0.03 --t=15.95 --K=20 --droprate=0.5

Cornell:

python3 main_mgc.py --mode='test' --seed=100 --dataset_config_path='./config/data/cornell.ini' --alpha=0.66 --t=38.49 --K=31 --droprate=0.6

Texas:

python3 main_mgc.py --mode='test' --seed=100 --dataset_config_path='./config/data/texas.ini' --alpha=0.75 --t=0.53 --K=4 --droprate=0.5

Washington:

python3 main_mgc.py --mode='test' --seed=100 --dataset_config_path='./config/data/washington.ini' --alpha=0.73 --t=42.36 --K=21 --droprate=0.1

Wisconsin:

python3 main_mgc.py --mode='test' --seed=100 --dataset_config_path='./config/data/wisconsin.ini' --alpha=0.34 --t=0.52 --K=12 --droprate=0.5

Releases

No releases published

Packages

No packages published