1: \begin{abstract}

2: Bundle methods have been intensively studied for solving both convex and nonconvex optimization problems. In most of the bundle methods developed thus far, at least one quadratic programming (QP) subproblem needs to be solved in each iteration. In this paper, we exploit the feasibility of developing a bundle algorithm that only solves linear subproblems. 

3: We start from minimization of a convex function and show that the sequence of major iterations converge to a minimizer. For nonconvex functions we  consider functions that are locally Lipschitz continuous and prox-regular on a bounded level set, and minimize the cutting-plane model over a trust region with infinity norm. The para-convexity of such functions allows us to use the locally convexified model and its convexity properties. Under some conditions and assumptions, we study the convergence of the proposed algorithm through the outer semicontinuity of the  proximal mapping. Encouraging results of preliminary numerical experiments on standard test sets are provided.

4: % suggest that conditions in the convergence analysis can be satisfied.

5: %algorithm is reliable and efficient.

6: %First we present a bundle trust region method for convex optimization which only solves linear subproblems. %sent a special bundle trust region method for minimizing locally Lipschitz and prox-regular functions. The model is to be controlled during the iteration process such that the linearization errors are always positive. The trust region is formed by infinity norm so we have a linear subproblem in each iteration. We show that if the convexification succeeds the algorithm converges to a stationary point. 

7: \end{abstract}