GraphFury 0.3: Expandable, concurrent learning on graphs with intuitive interfaces | AWS Machine Learning Blog

GraphStorm Overview

GraphStorm is a low-code enterprise graph machine learning (GML) framework designed to streamline the process of building, training, and deploying graph ML solutions on large enterprise-scale graphs within days, rather than months. It enables the creation of solutions that leverage the complex relationships and interactions between billions of entities in real-world data, such as fraud detection, recommendations, community detection, and search/retrieval problems.

GraphStorm 0.3 Features

The latest release of GraphStorm, version 0.3, introduces native support for multi-task learning on graphs. This update allows users to define multiple training targets on different nodes and edges within a single training loop. Additionally, new APIs have been added to customize GraphStorm pipelines, simplifying the process of implementing custom node classification training loops.

Multi-Task Learning on Graphs with GraphStorm 0.3

GraphStorm 0.3 supports multi-task learning on graphs across six common tasks: node classification, node regression, edge classification, edge regression, link prediction, and node feature reconstruction. Users can specify the training targets through a YAML configuration file, enabling the modeling of complex multi-task scenarios in enterprise applications.

Language Model Integration in GraphStorm

GraphStorm offers built-in techniques to efficiently train language models (LMs) and graph neural networks (GNNs) together on massive text-rich graphs. This integration allows for improved performance in scenarios where graphs contain text features, such as retail search applications. Customers can leverage pre-trained BERT+GNN or fine-tuned BERT+GNN methods to optimize model performance on text-rich datasets.

Scalability and Performance Benchmark

GraphStorm’s scalability is demonstrated through benchmarking on large synthetic graphs with billions of edges. The framework enables graph construction and model training on massive graphs, showcasing its ability to handle large-scale graph ML tasks efficiently.

Conclusion

GraphStorm 0.3, available under the Apache-2.0 license, empowers users to address large-scale graph ML challenges with enhanced multi-task learning capabilities, flexible APIs, and seamless integration of language models. Visit the GraphStorm GitHub repository and documentation to start leveraging these advanced features for your enterprise use cases.