dglai/KDD-2019-Hands-on
DGL tutorial in KDD 2019
| repo name | dglai/KDD-2019-Hands-on |
| repo link | https://github.com/dglai/KDD-2019-Hands-on |
| homepage | |
| language | Jupyter Notebook |
| size (curr.) | 20249 kB |
| stars (curr.) | 104 |
| created | 2019-06-21 |
| license | |
Learning Graph Neural Networks with Deep Graph Library – KDD'19 hands-on tutorial
Presenters: Minjie Wang, Lingfan Yu, Da Zheng, Nicholas Choma
Opening speaker: Alex Smola
Time: Wed, August 07, 2019, 1:30 pm - 4:30 pm
Abstract
Learning from graph data has played a substantial role in many real world scenarios including social network analysis, knowledge graph construction, protein function prediction and so on. Recent burst of researches on Graph Neural Networks (GNNs) brings representation learning to non-euclidean space and achieves state-of-art results in community detection, drug discovery, recommendation, etc. More recent perspective begins to view GNN as a more general form of the neural network models, such as attention architecture, that have dominated areas of computer vision and natural language processing. As graph is essentially relation, modeling explicit or inferring latent graph structure is crucial to the ability of relational reasoning for model AI.
This tutorial focuses on this recent trend in geometric deep learning including how and why graph neural networks are widely applied, its foundation and recent development. We then introduce a new framework called Deep Graph Library (DGL) that is designed to ease deep learning on graphs. The hands-on part starts with basic concepts in DGL for easier understanding, and later walks the audience through several end-to-end examples including community detection, hierarchical clustering and building recommender system using GNNs.
Prerequisite
Basic understanding of Machine Learning and Deep Learning. Have experience with either Pytorch or Apache MXNet.
Agenda
| Time | Session | Slides | Notebooks | Presenter |
|---|---|---|---|---|
| 1:30-2:00 | Opening talk | - | - | Alex Smola |
| 2:00-2:45 | DGL 101(Hands-on) Semi-supervised Community Detection using Graph Convolutional Network | link | link | Lingfan Yu |
| 2:45-3:30 | Scalable Clustering with Graph Neural Networks using DGL(Hands-on) GNNs for clustering TrackML dataset | link | link | Nicholas Choma / Minjie Wang |
| 3:30-3:45 | Coffee Break | |||
| 3:45-4:30 | Building Recommender Systems using Graph Neural Networks(Hands-on) GraphSage for MovieLens | link | link | Da Zheng |
Community
Join our Slack using this invitation link (expired in a week). Jumpy to the kdd19-tutorial channel for community meetup!
Play locally
Build a docker image with all the environment installed.
docker build --force-rm -t dgl-kdd19 -f Dockerfile .
Start a container using the image,
docker run -it --rm -p 8888:8888 dgl-kdd19 bash
Within the docker image,
cd ~/KDD-2019-Hands-on
conda activate kdd19
jupyter notebook --ip 0.0.0.0 --allow-root
Finally, open the url with browswer.
