Estimating the Cheeger constant using machine learning

This work addresses graph theory and machine learning applications by providing a novel computational method for estimating the Cheeger constant, which is incremental in combining existing techniques.

The paper tackled the problem of estimating the Cheeger constant of graphs by showing it has a predominant linear dependence on the largest two eigenvalues and using a trained deep neural network as an effective estimator, achieving results that generalize from smaller to larger graphs.