Graphsage

Tensorflow implementation of 'Inductive Representation Learning on Large Graphs'

Introduction

A tensorflow re-implementation of graphsage, which is easier than the original implementation GraphSAGE original implementation.
This code includes supervised and uinsupervised version, and three types of aggregators('mean','pooling' and 'lstm').

Requirement

python 3.6, tensorflow 1.12.0

Usage

To see and modify the parameters of graphsage, see config.py.
To run the codes, use:

python main.py

Results

Here shows accuracy of the supervised and unsupervised graphsage with 'mean' aggregator.

The supervised graphsage accuracy is 0.871

The unsupervised graphsage accuracy is 0.790