cluster-hgl's Introduction

Cluster-HGL

A PyTorch implementation of "" (Submitted)

Abstract

AbstractCluster-GCN, in average.

This repository provides a PyTorch implementation of Cluster-HGL as described in the submitted paper:

TO DO Mahmood Amintoosi, JAC, 2021

Requirements

The codebase is implemented in Python 3.7.11 on Google colab. package versions used for development are just below.

torch-scatter 		2.0.8
torch-sparse		0.6.11
torch-geometric		2.0.1
texttable		1.6.4
karateclub		1.2.1

Run on Google Colab

https://colab.research.google.com/github/mamintoosi/Cluster-HGL/blob/master/OverlappingClusterGCN.ipynb

Datasets

We used some of the citation network datasets, which are accessible from PyTorch-geometric

Model options

  --clustering-method   STR     Clustering method.             Default is `DANMF`.
  --cluster-number      INT     Number of clusters.            Default is 2x Number of dataset labels. 
  --seed                INT     Random seed.                   Default is 42.
  --epochs              INT     Number of training epochs.     Default is 10.
  --test-ratio          FLOAT   Training set ratio.            Default is 0.3.
  --learning-rate       FLOAT   Adam learning rate.            Default is 0.01.
  --dropout             FLOAT   Dropout rate value.            Default is 0.5.
  --layers              LST     Layer sizes.                   Default is [16, 16, 16]. 
  --membership-closeness FLOAT  WMC parameter					Default is 0.1
  --dataset-name		STR		Dataset Name					Default is Cora

This code is heavily borrowed from ClusterGCN

Recommend Projects