Moreau envelope and proximal-point methods under the lens of high-order regularization

Abstract: This paper is devoted to investigating the fundamental properties of high-order proximal operator (HOPE) and high-order Moreau envelope (HOME) in the nonconvex setting, meaning that the quadratic regularization ($p=2$) is replaced with a regularization with $p>1$. After studying several basic properties of HOPE and HOME, we investigate the differentiability and weak smoothness of HOME under $q$-prox-regularity $q\geq 2$ and $p$-calmness for $p \in (1,2]$ and $2 \leq p \leq q$. Further, we design of a high-order proximal-point algorithm (HiPPA) for which the convergence of the generated sequence to proximal fixed points is studied. Our results pave the way toward the high-order smoothing theory with $p>1$ that can lead to algorithmic developments in the nonconvex setting, where our numerical experiments of HiPPA on Nesterov-Chebyshev-Rosenbrock functions show the potential of this development for nonsmooth and nonconvex optimization.