Tensor methods in optimization

In this talk we present p-order methods for unconstrained minimization of convex functions that are p-times differentiable with Hölder continuous p-th derivatives. We establish worst-case complexity bounds for methods with and without acceleration. Some of these methods are "universal", that is, they do not require prior knowledge of the constants that define the smoothness level of the objective function. A lower complexity bound for this problem class is also obtained. This is a joint work with Yurii Nesterov (Université Catholique de Louvain).