GraphSnapShot: A System for Graph Machine Learning Acceleration

Dong Liu, Yanxuan Yu

Presentation: Virtual

Keywords: ML System; Graph Learning; Efficient System Design

Presenter Full Name: Dong Liu

TL;DR: A system for fast storage, retrieval, caching of Large-Scale Graph Learning.

Presenter Email: [e-mail]

Abstract:

We present \textbf{GraphSnapShot}, a caching framework for efficient graph learning for graph machine learning at large scale. By combining static sampling with dynamic on-the-fly updates, GraphSnapShot reduces memory usage and computation overhead. Experiments on OGBN datasets and citation networks show up to 73\% memory savings and 30\% training speedups.

Code is available at https://github.com/NoakLiu/GraphSnapShot.

Presenter Bio: Dong Liu is the founder of FastLM.ai. He has been a graduate researcher at Yale University and he received his bachelor degree from University of Wisconsin-Madison, both in Computer Science. His interests lies in Fast Solution for Large Models (FastLM).

Paper Checklist Guidelines: I certify that all co-authors have validated the presented results and conclusions, and have read and commit to adhering to the Paper Checklist Guidelines, Call for Papers and Publication Ethics.

YouTube Link: https://youtu.be/Pf1CiZvXYo0

YouTube Link Poster: https://youtu.be/Pf1CiZvXYo0

Dataset Release: I certify that all co-authors commit to release the dataset and necessary scripts to reproduce the presented results.

Google Slides: https://drive.google.com/file/d/1z1c70gqUNZsynq6mVmT8m4FsxVdq_prS/view?usp=sharing

Poster: No

Workshop Registration: Yes, the presenter has registered for the workshop.

YouTube Link Short: In Recording

Submission Number: 12