Estimating the Cheeger constant using machine learning

05/12/2020

∙

by   Ambar Jain, et al.

∙

0

∙

share

In this paper, we use machine learning to show that the Cheeger constant of a connected regular graph has a predominant linear dependence on the largest two eigenvalues of the graph spectrum. We also show that a trained deep neural network on graphs of smaller sizes can be used as an effective estimator in estimating the Cheeger constant of larger graphs.