A bundle interior proximal method for solving convex minimization problems

In this paper we extend the standard bundle proximal method for finding the minimum of a convex not necessarily differentiable function on the nonnegative orthant. The strategy consists in approximating the objective function by a piecewise linear convex function and using distance-like functions based on second order homogeneous kernels. First we prove the convergence of this new bundle interior proximal method under the same assumptions as for the standard bundle method and then we report some preliminary numerical experiences for a particular distance function.