Accelerations for a variety of global optimization methods (Q1315439)

This paper studies acceleration techniques for a class of deterministic algorithms for global optimization. The acceleration techniques are applicable if the functions to be minimized have certain smoothness properties. They use the Lipschitz constant of the function and derivative information to construct better lower envelopes for the function. The acceleration techniques are derived from a geometric viewpoint. Numerical tests are performed, which compare the original algorithm and its accelerated version.