Graph-structured data is omnipresent in the real world. With their capacity to model intricate relationships between entities, graphs serve as a fundamental data structure for representing complex structured knowledge, spanning domains like social networks, chemical structures, knowledge graphs, and transportation networks. In recent years, the study of graph machine learning techniques, such as network embedding methods and Graph Neural Networks, has garnered substantial attention. State-of-the-art methods have exhibited remarkable performance across a spectrum of tasks such as node classification, link prediction, and graph classification. These tasks underpin numerous practical applications, including recommendation systems, predictive user behavior modeling, and drug discovery.

Despite the undeniable success of graph machine learning techniques on standardized benchmarks, their adoption in real industrial applications remains limited, primarily due to scalability challenges. Real-world graphs are characterized by their sheer size and the interconnected nature of the data, posing formidable obstacles for engineers and researchers striving to leverage graph machine learning. For instance, in platforms with a billion-scale user population like Meta, Amazon, Snapchat, etc, the graph data may even surpass the storage capacity of a single machine, making efficient graph machine learning model training a complex endeavor that demands meticulous modeling design and engineering effort. Additionally, the non-i.i.d. nature of graph data introduces further obstacles in real-time inference with graph data, including neighbor explosion and high latency in neighborhood fetching.

This Research Topic aims to foster collaboration between academic and industrial researchers from diverse backgrounds and perspectives to address the pressing challenges in scaling graph machine learning for real-world applications. We invite submissions that explore innovative solutions, novel methodologies, and practical insights into overcoming scalability constraints, making graph machine learning more accessible and impactful in industrial settings.

Submissions are encouraged in, but not limited to, the following areas:

● Scalable algorithms and techniques for graph machine learning

● Distributed and parallel computing approaches for handling and training with large-scale graph data

● Case studies and success stories of applying graph machine learning in real industrial applications

● Benchmark datasets and evaluation metrics that reflect real-world challenges

● Scalability-aware model architectures and design principles

● Strategies for handling non-i.i.d. graph data and mitigating neighbor explosion

● Efficient training and/or inference for graph machine learning

Topic Editor Tong Zhao is employed by Snap Inc. All other Topic Editors declare no competing interests with regards to the Research Topic subject.

Submit your idea

Impact

Frontiers in Big Data

Submit your idea

Impact