Revolutionizing Graph Machine Learning: Introducing GraphStorm 0.3
GraphStorm is a cutting-edge low-code enterprise graph machine learning (GML) framework that empowers users to build, train, and deploy graph ML solutions on complex enterprise-scale graphs in a matter of days, rather than months. With GraphStorm, you have the ability to create solutions that leverage the intricate relationships and interactions between billions of entities present in real-world data. From fraud detection to recommendations, community detection, and search/retrieval problems, GraphStorm offers a versatile platform for solving a wide range of challenges.
Native Support for Multi-Task Learning on Graphs
Today, we are thrilled to announce the launch of GraphStorm 0.3, which introduces native support for multi-task learning on graphs. This feature enables users to define and train multiple targets on different nodes and edges within a single training loop. With the addition of new APIs, customizing GraphStorm pipelines has become simpler than ever. Just 12 lines of code are all you need to implement a custom node classification training loop. To help you get started, we have included two Jupyter notebook examples for node classification and link prediction tasks.
New APIs for Customizing GraphStorm Pipelines
GraphStorm 0.3 takes customization to the next level with new APIs that allow users to tailor their training and inference pipelines to meet specific requirements. Whether you’re a seasoned data scientist or a beginner in the field of graph ML, these APIs streamline the process of building and deploying models. By integrating these APIs, you can efficiently define custom node classification training pipelines with ease.
Comprehensive Study of LM+GNN for Large Text-Rich Graphs
In enterprise applications where text features play a critical role, GraphStorm shines with its ability to train language models (LMs) and GNN models together efficiently on massive text-rich graphs. To demonstrate the power of this approach, we conducted a benchmark study using the Microsoft Academic Graph (MAG) dataset. By combining pre-trained BERT models with GNNs, GraphStorm delivers exceptional performance on tasks like node classification and link prediction. The study showcases the scalability and effectiveness of GraphStorm in handling large text-rich graphs.
Conclusion
GraphStorm 0.3 is a game-changer in the world of graph machine learning, offering unparalleled capabilities for tackling large-scale graph ML challenges. With native support for multi-task learning, customizable pipelines, and advanced LM+GNN techniques, GraphStorm is the go-to framework for complex graph analytics. To get started with GraphStorm, visit our GitHub repository and documentation to explore the endless possibilities it offers.
About the Authors
Xiang Song is a senior applied scientist at AWS AI Research and Education (AIRE), with a background in developing deep learning frameworks including GraphStorm. Jian Zhang and Florian Saupe are seasoned applied scientists with extensive experience in machine learning and graph neural networks. Their contributions to the graph machine learning community have been invaluable.