On Efficiency of Tightening Bounds in Interval Global Optimization

摘要

The tightness of bounds is very important factor of efficiency of branch and bound based global optimization algorithms. An experimental model of interval arithmetic with controllable tightness of bounds is proposed to investigate the impact of bounds tightening in interval global optimization. A parallel version of the algorithm is implemented to cope with the computational intensity of the experiment. The experimental results on efficiency of tightening bounds are presented for several test and practical problems. The suitability of mater-slave type parallelization to speed up the experiments is justified.