Timezone: »
Google Meet Link: https://meet.google.com/yko-cpuk-czg
Slides: https://drive.google.com/file/d/1ECcnRgJqjmj7hlegYuPdscLCUw7YJc7G/view?usp=sharing
Motivation
Graphs are general data structures that can represent information from a variety of domains (social, biomedical, online transactions, and many more). Graph Neural Networks (GNNs) are an exciting way to use graph structured data inside neural network models that have recently exploded in popularity. However, implementing GNNs and running GNNs on large (and complex) datasets still raises a number of challenges for machine learning platforms.
Goals
The main goal of this tutorial is to help practitioners and researchers to implement GNNs in a TensorFlow setting. Specifically, the tutorial will be mostly hands-on, and will walk the audience through a process of running existing GNNs on heterogeneous graph data, and a tour of how to implement new GNN models. The hands-on portion of the tutorial will be based on TF-GNN, a new framework that we open-sourced.
Learning objectives
1. Conceptual understanding of Graph Neural Networks (GNNs).
2. Hands-on: How to train and evaluate GNNs in TensorFlow, using TF-GNN.
3. Understanding of message passing building blocks for crafting advanced GNN architectures.
4. Hands-on: How to implement custom models inside TF-GNN.
5. Know how to run TF-GNN models at scale, using cloud environments.
6. Hands-on: Run TF-GNN at scale on large graphs.
Structure
This tutorial consists of 3 lectures, paired with 3 python notebooks, which cover different aspects of working with TF-GNN.
- 9:30 AM. Basics of TF-GNN
- 10:30 AM. Modeling with TF-GNN
- 11:30 AM. Scaling GNNs w/ TF-GNN
Material
ALL Presentation Slides
The notebooks (and recordings eventually) can be found on the tutorial website
Additional Resources
If you're interested in learning more about GNNs or TF-GNN, we recommend the following resources:
- Our paper TF-GNN: Graph Neural Networks in TensorFlow, details the API design and background of the library.
- The in-depth notebook OGBN-MAG end-to-end with TF-GNN offers a deep dive on building heterogeneous graph models using TF-GNN.
Mon 7:30 a.m. - 7:40 a.m.
|
Welcome
(
Introduction
)
|
Bryan Perozzi 🔗 |
Mon 7:40 a.m. - 8:00 a.m.
|
GNN Basics
(
Talk
)
A brief overview of GNNs and motivation for TF-GNN. |
Sami A Abu-El-Haija 🔗 |
Mon 8:00 a.m. - 8:20 a.m.
|
TF-GNN Basics (Hands on)
(
Demonstration
)
Overview of running TF-GNN models on small scale, in-memory datasets. |
Sami A Abu-El-Haija 🔗 |
Mon 8:20 a.m. - 8:30 a.m.
|
Break
|
🔗 |
Mon 8:30 a.m. - 8:50 a.m.
|
TF-GNN Modeling
(
Talk
)
A more detailed dive into data representation and modeling primitives available in TF-GNN. |
Neslihan Bulut 🔗 |
Mon 8:50 a.m. - 9:20 a.m.
|
TF-GNN Modeling (Hands on)
(
Demonstration
)
Hands-on walk through a notebook illustrating how to use and create a TF-GNN model. |
Neslihan Bulut 🔗 |
Mon 9:20 a.m. - 9:30 a.m.
|
Break
|
🔗 |
Mon 9:30 a.m. - 9:50 a.m.
|
Running TF-GNN at Scale
(
Talk
)
Covers how to run TF-GNN at scale, covering: 1) Common ML system architectures 2) Batch architectures for scaling Graph Neural Networks 3) Introduction to Apache Beam for Scalable Graph Sub-Sampling |
Brandon Mayer 🔗 |
Mon 9:50 a.m. - 10:20 a.m.
|
Running TF-GNN at Scale (Hands on)
(
Demonstration
)
Hands-on walk through a notebook illustrating using TF-GNN with cloud computing. Covers distributed sampling, and model training with VertexAI. |
Brandon Mayer 🔗 |
Mon 10:20 a.m. - 10:25 a.m.
|
Closing
|
Anton Tsitsulin 🔗 |
Author Information
Bryan Perozzi (Google Research)
Sami A Abu-El-Haija (Google Research)
Neslihan Bulut (Google)
Brandon Mayer (Google)
More from the Same Authors
-
2022 : GraphWorld: Fake Graphs BringReal Insights for GNNs »
John Palowitch · Anton Tsitsulin · Bryan Perozzi · Brandon Mayer -
2022 : Differentially Private Graph Learning via Sensitivity-Bounded Personalized PageRank »
Alessandro Epasto · Vahab Mirrokni · Bryan Perozzi · Anton Tsitsulin · Peilin Zhong -
2023 Poster: Learning Large Graph Property Prediction via Graph Segment Training »
Kaidi Cao · Phitchaya Phothilimtha · Sami Abu-El-Haija · Dustin Zelle · Yanqi Zhou · Charith Mendis · Jure Leskovec · Bryan Perozzi -
2023 Poster: TpuGraphs: A Performance Prediction Dataset on Large Tensor Computational Graphs »
Phitchaya Phothilimtha · Sami Abu-El-Haija · Kaidi Cao · Bahare Fatemi · Charith Mendis · Bryan Perozzi -
2023 Poster: Graph Clustering with Graph Neural Networks »
Anton Tsitsulin · John Palowitch · Bryan Perozzi · Emmanuel Müller -
2022 Spotlight: Machine Learning on Graphs: A Model and Comprehensive Taxonomy »
Ines Chami · Sami Abu-El-Haija · Bryan Perozzi · Christopher Ré · Kevin Murphy -
2022 Poster: Differentially Private Graph Learning via Sensitivity-Bounded Personalized PageRank »
Alessandro Epasto · Vahab Mirrokni · Bryan Perozzi · Anton Tsitsulin · Peilin Zhong -
2022 Poster: Zero-shot Transfer Learning within a Heterogeneous Graph via Knowledge Transfer Networks »
Minji Yoon · John Palowitch · Dustin Zelle · Ziniu Hu · Ruslan Salakhutdinov · Bryan Perozzi -
2022 Poster: Machine Learning on Graphs: A Model and Comprehensive Taxonomy »
Ines Chami · Sami Abu-El-Haija · Bryan Perozzi · Christopher Ré · Kevin Murphy -
2022 : Running TF-GNN at Scale (Hands on) »
Brandon Mayer -
2022 : Running TF-GNN at Scale »
Brandon Mayer -
2022 : TF-GNN Modeling (Hands on) »
Neslihan Bulut -
2022 : TF-GNN Modeling »
Neslihan Bulut -
2022 : TF-GNN Basics (Hands on) »
Sami A Abu-El-Haija -
2022 : GNN Basics »
Sami A Abu-El-Haija -
2022 : Welcome »
Bryan Perozzi -
2021 Poster: Shift-Robust GNNs: Overcoming the Limitations of Localized Graph Training data »
Qi Zhu · Natalia Ponomareva · Jiawei Han · Bryan Perozzi -
2020 : GNNs and Graph Embeddings »
Bryan Perozzi -
2020 Expo Workshop: Mining and Learning with Graphs at Scale »
Vahab Mirrokni · Bryan Perozzi · Jakub Lacki · Jonathan Halcrow · Jaqui C Herman