Graph Bayesian Optimization: Algorithms, Evaluations and Applications

TL;DR

This paper introduces a flexible graph Bayesian optimization framework that leverages graph kernels and structural features to optimize complex network structures, demonstrating its effectiveness through multiple evaluations and applications.

Contribution

It presents a novel framework for Bayesian optimization on arbitrary graphs, integrating graph kernels and feature importance analysis, which was not addressed in prior work.

Findings

Effective optimization of complex network structures.

Framework identifies important features during optimization.

Demonstrated success in multiple evaluation and application scenarios.

Abstract

Network structure optimization is a fundamental task in complex network analysis. However, almost all the research on Bayesian optimization is aimed at optimizing the objective functions with vectorial inputs. In this work, we first present a flexible framework, denoted graph Bayesian optimization, to handle arbitrary graphs in the Bayesian optimization community. By combining the proposed framework with graph kernels, it can take full advantage of implicit graph structural features to supplement explicit features guessed according to the experience, such as tags of nodes and any attributes of graphs. The proposed framework can identify which features are more important during the optimization process. We apply the framework to solve four problems including two evaluations and two applications to demonstrate its efficacy and potential applications.