In this talk, I will present the mean-field analysis framework for overparameterized Deep ResNets. I will first establish the rigorous convergence of the finite-width training dynamics to their mean-field limit and then discuss an application of this limit process that demonstrates the convergence of the training dynamics.