An Example of Only Linear Convergence of Trust Region Algorithms for Non-smooth Optimization
Y. Yuan · IMA Journal of Numerical Analysis · 1984
Most superlinear convergence results about trust region algorithms for non-smooth optimization are dependent on the inactivity of trust region restrictions. An example is constructed to show that it is possible that at every iteration the trust region bound is active and the rate of convergence is only linear, though strict complementarity and second order sufficiency conditions are satisfied.