Generalization Guarantees for Neural Architecture Search with
                 Train-Validation Split

                 Samet Oymak 1 Mingchen Li 2 Mahdi Soltanolkotabi 3

             Abstract                 portant for deep learning applications where there are many
    Neural Architecture Search (NAS) is a popular        possibilities for choosing a variety of hyperparameters to
    method for automatically designing optimized         achieve the best test accuracy. A crucial application of  ...