Response Surfaces with Discounted Information for Global Optima Tracking in Dynamic Environments

Two new methods for incorporating old and recent information into a surrogate model in order to improve the tracking of the global optima of expensive black boxes are presented in this paper. The response surfaces are built using Gaussian processes fitted to data which is obtained through sequential sampling. The efficient global optimization (EGO) algorithm applied to the generated response surface is used to determine the next most promising sample (where the expected improvement is maximized). The goal is to find the global maxima of an expensive to evaluate objective function which changes after a given number of function evaluations with as few samples as possible. Exploiting old information in a discounted manner significantly improves the search, which is shown through numerical experiments performed using the moving peaks benchmark (MPB).

[1]  Rolf Drechsler,et al.  Applications of Evolutionary Computing, EvoWorkshops 2008: EvoCOMNET, EvoFIN, EvoHOT, EvoIASP, EvoMUSART, EvoNUM, EvoSTOC, and EvoTransLog, Naples, Italy, March 26-28, 2008. Proceedings , 2008, EvoWorkshops.

[2]  Tobias Wagner,et al.  Model-based optimization revisited: Towards real-world processes , 2008, 2008 IEEE Congress on Evolutionary Computation (IEEE World Congress on Computational Intelligence).

[3]  Jianqiao Chen,et al.  A surrogate-based particle swarm optimization algorithm for solving optimization problems with expensive black box functions , 2013 .

[4]  Hendrik Richter,et al.  Detecting change in dynamic fitness landscapes , 2009, 2009 IEEE Congress on Evolutionary Computation.

[5]  Michael N. Vrahatis,et al.  Unified Particle Swarm Optimization in Dynamic Environments , 2005, EvoWorkshops.

[6]  Donald R. Jones,et al.  Efficient Global Optimization of Expensive Black-Box Functions , 1998, J. Glob. Optim..

[7]  Manfred Jaeger,et al.  Proceedings of the 24th Annual International Conference on Machine Learning (ICML 2007) , 2007, ICML 2007.

[8]  Carl E. Rasmussen,et al.  Gaussian processes for machine learning , 2005, Adaptive computation and machine learning.

[9]  Donald R. Jones,et al.  A Taxonomy of Global Optimization Methods Based on Response Surfaces , 2001, J. Glob. Optim..

[10]  Jürgen Branke,et al.  Memory enhanced evolutionary algorithms for changing optimization problems , 1999, Proceedings of the 1999 Congress on Evolutionary Computation-CEC99 (Cat. No. 99TH8406).

[11]  Christodoulos A. Floudas,et al.  A review of recent advances in global optimization , 2009, J. Glob. Optim..

[12]  Shigeyoshi Tsutsui,et al.  Advances in evolutionary computing: theory and applications , 2003 .

[13]  G. Gary Wang,et al.  Survey of modeling and optimization strategies to solve high-dimensional design problems with computationally-expensive black-box functions , 2010 .

[14]  Hartmut Schmeck,et al.  Designing evolutionary algorithms for dynamic optimization problems , 2003 .

[15]  Jürgen Branke,et al.  Evolutionary optimization in uncertain environments-a survey , 2005, IEEE Transactions on Evolutionary Computation.

[16]  Michael C. Burl,et al.  Active learning for directed exploration of complex systems , 2009, ICML '09.

[17]  Jürgen Branke,et al.  Multiswarms, exclusion, and anti-convergence in dynamic environments , 2006, IEEE Transactions on Evolutionary Computation.

[18]  Jürgen Branke,et al.  Revenue maximization through dynamic pricing under unknown market behaviour , 2012, SCOR.

[19]  Russell C. Eberhart,et al.  Adaptive particle swarm optimization: detection and response to dynamic systems , 2002, Proceedings of the 2002 Congress on Evolutionary Computation. CEC'02 (Cat. No.02TH8600).

[20]  Anabela Simões,et al.  Improving memory’s usage in evolutionary algorithms for changing environments , 2007, 2007 IEEE Congress on Evolutionary Computation.