论文信息 - Black-box optimization benchmarking of NEWUOA compared to BIPOP-CMA-ES: on the BBOB noiseless testbed

Black-box optimization benchmarking of NEWUOA compared to BIPOP-CMA-ES: on the BBOB noiseless testbed

In this paper, the performances of the NEW Unconstrained Optimization Algorithm (NEWUOA) on some noiseless functions are compared to those of the BI-POPulation Covariance Matrix Adaptation-Evolution Strategy (BIPOP-CMA-ES). The two algorithms were benchmarked on the BBOB 2009 noiseless function testbed. The comparison shows that NEWUOA outperforms BIPOP-CMA-ES on some functions like the Sphere or the Rosenbrock functions. Also the independent restart procedure used for NEWUOA allows it to perform better than BIPOP-CMA-ES on the Gallagher functions. Nevertheless, BIPOP-CMA-ES is faster and has a better success probability than NEWUOA in reaching target function values smaller than one on all other functions.

Raymond Ros | Nikolaus Hansen | N. Hansen | Raymond Ros

[1] Anne Auger,et al. Empirical comparisons of several derivative free optimization algorithms , 2009 .

[2] Raymond Ros,et al. Real-Parameter Black-Box Optimization Benchmarking 2009: Experimental Setup , 2009 .

[3] Edmund K. Burke,et al. The Genetic and Evolutionary Computation Conference , 2011 .

[4] M. Powell. The NEWUOA software for unconstrained optimization without derivatives , 2006 .

[5] K. Price. Differential evolution vs. the functions of the 2/sup nd/ ICEO , 1997, Proceedings of 1997 IEEE International Conference on Evolutionary Computation (ICEC '97).

[6] Raymond Ros,et al. Benchmarking the NEWUOA on the BBOB-2009 function testbed , 2009, GECCO '09.

[7] Nikolaus Hansen,et al. Benchmarking a BI-population CMA-ES on the BBOB-2009 function testbed , 2009, GECCO '09.

[8] Anne Auger,et al. Experimental Comparisons of Derivative Free Optimization Algorithms , 2009, SEA.

[9] Anne Auger,et al. Real-Parameter Black-Box Optimization Benchmarking 2009: Noiseless Functions Definitions , 2009 .