GitHub GitHub last commit Own Badge Own Badge
By integrating the DGL framework with the Neo4j DBMS, analysis of network data is more wholistic, combining the visualization prowess and "classical" graph algorithms of Neo4j and the end-to-end deep learning models of DGL.
Overall, this repository not only solves both Neo4j and DGL's shortcomings, but also provides visualizations of model training for enhanced understanding of GCN/GNNs.
Project Goals:
This repository was created with the goal in mind of integrating both DGL and neo4j to provide a complete "stack" of network analysis tools, and then use these tools for analysis of social networks.
DataRes Network:
The social network that is used is DataRes @ UCLA, a student organization consisting of data science teams with a variety of goals.
The Research Team studies deep learning applications, the Athletics team uses data science and statistics to aid UCLA Athletics, the Consulting team helps private organizations pro-bono, and the Data Blog team analyzes interesting topics in society!
More Info: https://ucladatares.com
Requirements:
This repository is intended to be used in conjunction with Neo4j, a graph application found here.
After downloading Neo4j, the DBMSS should contain both APOC and GDSL. It is only then will this repository run cleanly.
The code to deploy this graph is in the folder /graph_create/
cd graph_create
python3 graph_create.py --uri [URI] --password [PASSWORD]
It is best practice to create a new conda environment to run this repos.
conda create -n dgl
conda activate dgl
git clone git@github.com:datares/neodgl.git
pip install -r requirements.txt
Start Training with Zach's Karate Club Test Data:
python3 train.py --test
To use your own database:
python3 train.py --uri [DBMS URI] --password [DBMS PASSWORD]
Arguments:
usage: train.py [-h] [--test] [--uri URI] [--password PASSWORD]
Train the GCN model using either the test data or neo4j dbms
optional arguments:
-h, --help show this help message and exit
--test train the test data!
--uri URI neo4j dbms uri
--password PASSWORD neo4j dbms password
Semi-Supervised Node Classification Model Training:
