Umut Simsekli

We prove algorithm- and data-dependent upper bounds on the generalization error of diffusion models by using tools from statistical learning theory