Deep Learning with Kernels through RKHM and the Perron-Frobenius Operator Tokyo, Japan

We derive a new Rademacher generalization bound in this setting and provide a theoretical interpretation of benign overfitting by means of Perron-Frobenius operators.