Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Towards Versatile Graph Learning Approach: from the Perspective of Large Language Models

Lanning Wei, Jun Gao, Huan Zhao, Quanming Yao

TL;DR

Graph learning across diverse tasks and domains remains challenging due to varied data and procedure complexity. The authors propose a conceptual prototype that uses large language models to address the 'where' and 'how' of graph learning across four pipeline stages: task definition, feature engineering, model selection and optimization, and deployment and serving. They review existing methods, classify them by graph data types (textual, structured-textual, and structured) and LLM roles (predictor, co-operator, advisor), and outline future directions including graph foundation models and universal graph learning agents. The work offers a holistic framework to leverage LLMs for versatile, explainable, and scalable graph learning across domains.

Abstract

Graph-structured data are the commonly used and have wide application scenarios in the real world. For these diverse applications, the vast variety of learning tasks, graph domains, and complex graph learning procedures present challenges for human experts when designing versatile graph learning approaches. Facing these challenges, large language models (LLMs) offer a potential solution due to the extensive knowledge and the human-like intelligence. This paper proposes a novel conceptual prototype for designing versatile graph learning methods with LLMs, with a particular focus on the "where" and "how" perspectives. From the "where" perspective, we summarize four key graph learning procedures, including task definition, graph data feature engineering, model selection and optimization, deployment and serving. We then explore the application scenarios of LLMs in these procedures across a wider spectrum. In the "how" perspective, we align the abilities of LLMs with the requirements of each procedure. Finally, we point out the promising directions that could better leverage the strength of LLMs towards versatile graph learning methods.

Towards Versatile Graph Learning Approach: from the Perspective of Large Language Models

TL;DR

Graph learning across diverse tasks and domains remains challenging due to varied data and procedure complexity. The authors propose a conceptual prototype that uses large language models to address the 'where' and 'how' of graph learning across four pipeline stages: task definition, feature engineering, model selection and optimization, and deployment and serving. They review existing methods, classify them by graph data types (textual, structured-textual, and structured) and LLM roles (predictor, co-operator, advisor), and outline future directions including graph foundation models and universal graph learning agents. The work offers a holistic framework to leverage LLMs for versatile, explainable, and scalable graph learning across domains.

Abstract

Graph-structured data are the commonly used and have wide application scenarios in the real world. For these diverse applications, the vast variety of learning tasks, graph domains, and complex graph learning procedures present challenges for human experts when designing versatile graph learning approaches. Facing these challenges, large language models (LLMs) offer a potential solution due to the extensive knowledge and the human-like intelligence. This paper proposes a novel conceptual prototype for designing versatile graph learning methods with LLMs, with a particular focus on the "where" and "how" perspectives. From the "where" perspective, we summarize four key graph learning procedures, including task definition, graph data feature engineering, model selection and optimization, deployment and serving. We then explore the application scenarios of LLMs in these procedures across a wider spectrum. In the "how" perspective, we align the abilities of LLMs with the requirements of each procedure. Finally, we point out the promising directions that could better leverage the strength of LLMs towards versatile graph learning methods.
Paper Structure (20 sections, 4 figures, 1 table)

This paper contains 20 sections, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Overview
  3. Machine Learning on Graphs
  4. Conceptual Prototype
  5. Comparisons with Contemporaneous Surveys
  6. Graph Learning with LLMs
  7. Task Definition
  8. Feature Engineering
  9. Textual Feature Construction
  10. Textual Feature Encoding
  11. Summary and Discussion
  12. Model Selection and Optimization
  13. LLMs as Predictor for Textual Graph $\mathcal{G}^T$
  14. LLMs as Co-operator for Structured-Textual Graphs $\mathcal{G}^{ST}$
  15. LLMs as Advisor for Structured graph $\mathcal{G}^S$
  16. ...and 5 more sections

Figures (4)

  • Figure 1: The conceptual prototype of versatile graph learning methods joint with LLMs. LLMs can be used in sequential graph learning procedures in these columns, with increased requirements for LLMs in different rows for each procedure. The rows can be further developed along with the exploration of different abilities of LLMs.
  • Figure 2: Illustrations of graph data feature engineering strategies that jointed with LLMs.
  • Figure 3: Illustrations of LLM-related graph learning algorithms on structured-textual graphs.
  • Figure 4: Illustrations of LLM-empowered graph learning algorithms on graph-structured data.