A Review of combinatorial optimization with graph neural networks

Abstract

In the last two decades, research work on neural networks have been shown successful in a number of domains, but due to the poor interpretability of neural networks, the research work on neural networks has not received much attention and attention in this century. However, the success of graph neural networks has boosted research on combinatorial optimization in these years. This greatly stimulated the enthusiasm of the researchers, resulting in a series of outcomes related to combinatorial optimization. In the paper, We divide the related methods into three, graph networks, combined with classical algorithms, combined with machine learning. We also analyze the differences of these methods. Finally, we briefly outline their applications and discuss potential future directions.