We apply the idea of infeasible bundle method to nonsmooth optimization problem based on CVaR (conditional value-at-risk) portfolio. The subproblem for generating next iterate point is considered by means of improvement function. With the help of Lagrangian function and dual theory, the explicit expression of next trial point for infeasible bundle method is obtained. Compared with  the available results, our method looses the requirement of feasibility at each iteration, which makes the method more direct and easier to implement.

nonsmooth optimization, CVaR (conditional value-at-risk), infeasible bundle method, subgradient.