Summary of Generalizing Graph Transformers Across Diverse Graphs and Tasks Via Pre-training on Industrial-scale Data, by Yufei He et al.
Generalizing Graph Transformers Across Diverse Graphs and Tasks via Pre-Training on Industrial-Scale Data
by Yufei He, Zhenyu Hou, Yukuo Cen, Feng He, Xu Cheng, Bryan Hooi
First submitted to arxiv on: 4 Jul 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Social and Information Networks (cs.SI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed PGT (Pre-trained Graph Transformer) framework is a scalable transformer-based graph pre-training model that can handle web-scale graphs with billions of nodes in industrial scenarios. The model aims to develop an inductive ability to make predictions for unseen new nodes and even new graphs. To achieve this, the authors design a flexible and scalable graph transformer as the backbone network, paired with two pre-training tasks based on the masked autoencoder architecture: reconstructing node features and local structures. Unlike traditional autoencoders, PGT utilizes the decoder for feature augmentation. The framework is deployed on Tencent’s online game data and demonstrates effective generalization to unseen new graphs with different downstream tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary PGT is a new way to train graph models that can handle very large graphs like those found in real-world applications. It’s designed to be fast and efficient, so it can be used for big datasets. The model has two main parts: a transformer backbone and two pre-training tasks. One task helps the model learn about node features, while the other helps it understand local graph structures. PGT is tested on Tencent’s game data and does well, even when applied to new graphs with different tasks. |
Keywords
» Artificial intelligence » Autoencoder » Decoder » Generalization » Transformer
