×

A hybrid model-classifier framework for managing prediction uncertainty in expensive optimisation problems. (English) Zbl 1307.90208

Summary: Many real-world optimisation problems rely on computationally expensive simulations to evaluate candidate solutions. Often, such problems will contain candidate solutions for which the simulation fails, for example, due to limitations of the simulation. Such candidate solutions can hinder the effectiveness of the optimisation since they may consume a large portion of the optimisation budget without providing new information to the optimiser, leading to search stagnation and a poor final result. Existing approaches to handle such designs either discard them altogether, or assign them a penalised fitness. However, this results in loss of beneficial information, or in a model with a severely deformed landscape. To address these issues, this study proposes a hybrid classifier-model framework. The role of the classifier is to predict which candidate solutions are likely to crash the simulation, and this prediction is then used to bias the search towards valid solutions. Furthermore, the proposed framework employs a trust-region approach, and several other procedures, to manage the model and classifier, and to ensure the progress of the optimisation. Performance analysis using an engineering application of airfoil shape optimisation shows the efficacy of the proposed framework, and the possibility to use the knowledge accumulated in the classifier to gain new insights into the problem being solved.

MSC:

90C59 Approximation methods and heuristics in mathematical programming
93C41 Control/observation systems with incomplete information

Software:

XFOIL; GAToolBox
PDFBibTeX XMLCite
Full Text: DOI

References:

[1] Bishop CM, Neural Networks for Pattern Recognition (1995)
[2] DOI: 10.1109/TSMCC.2004.841917 · doi:10.1109/TSMCC.2004.841917
[3] Buhmann MD, in Cambridge Monographs on Applied and Computational Mathematics 12 (2003)
[4] DOI: 10.1162/089976603321891864 · Zbl 1046.62001 · doi:10.1162/089976603321891864
[5] Chipperfield, A. Fleming, P., Pohlheim, H., and Fonseca, C. (1994),Genetic Algorithm TOOLBOX For Use with MATLAB, Version 1.2, Department of Automatic Control and Systems Engineering, University of Sheffield, Sheffield
[6] DOI: 10.1137/1.9780898719857 · Zbl 0958.65071 · doi:10.1137/1.9780898719857
[7] Conn AR, Approximation Theory and Optimization: Tributes to M.J.D. Powell pp 83– (1997)
[8] Conn AR, in Proceedings of the Seventh AIAA/USAF/NASA/ISSMO Symposium on Multidisciplinary Analysis and Optimization (1998)
[9] Cressie NAC, Statistics for Spatial Data (1993)
[10] Drela, M. and Youngren, H. (2001),XFOIL 6.9 User Primer, Department of Aeronautics and Astronautics, Massachusetts Institute of Technology, Cambridge, MA
[11] Duda RO, Pattern Classification ,, 2. ed. (2001)
[12] Emmerich, MTM. Giotis, A. Özedmir, M., Bäck, T., and Giannakoglou, K.C. (2002), ’Metamodel-assisted Evolution Strategies’, inThe 7th International Conference on Parallel Problem Solving from Nature–PPSN VII, ed. J.J. Merelo Guervós, no. 2439 in Lecture Notes in Computer Science, Berlin: Springer, pp. 361–370
[13] DOI: 10.1109/TEVC.2009.2039141 · Zbl 05921187 · doi:10.1109/TEVC.2009.2039141
[14] DOI: 10.1109/TEVC.2002.800884 · Zbl 05452037 · doi:10.1109/TEVC.2002.800884
[15] DOI: 10.1023/A:1008306431147 · Zbl 0917.90270 · doi:10.1023/A:1008306431147
[16] Koehler JR, Handbook of Statistics pp 261– (1996)
[17] DOI: 10.1109/TMTT.2006.882894 · doi:10.1109/TMTT.2006.882894
[18] Liang KH, International Journal of Knowledge-Based Intelligent Engineering Systems 4 pp 172– (2000)
[19] MacQueen, JB. (1967), ’Some Methods for Classification and Analysis of Multivariate Observations’, inProceedings of 5thBerkeley Symposium on Mathematical Statistics and Probability, eds. J. Neyman and L.M. Le Cam, Berkeley: University of California Press, pp. 281–297
[20] DOI: 10.1016/0898-1221(92)90175-H · Zbl 0766.41003 · doi:10.1016/0898-1221(92)90175-H
[21] DOI: 10.1080/00401706.1979.10489755 · doi:10.1080/00401706.1979.10489755
[22] Michalewicz Z, How to Solve It: Modern Heuristics (2004) · doi:10.1007/978-3-662-07807-5
[23] Myers RH, Response Surface Methodology: Process and Product Optimization Using Designed Experiments (1995) · Zbl 1161.62392
[24] DOI: 10.1108/03321640810861043 · Zbl 1144.78335 · doi:10.1108/03321640810861043
[25] Ong, YS. Nair, P.B., Keane, A.J., and Wong, K.W. (2005), ’Surrogate-assisted Evolutionary Optimization Frameworks for High-fIdelity Engineering Design Problems’, inKnowledge Incorporation in Evolutionary Computation, ed. Y. Jin, no. 167 in Studies in Fuzziness and Soft Computing, Berlin; New York: Springer, pp. 307–332
[26] DOI: 10.1016/S0045-7825(99)00394-1 · Zbl 0956.76023 · doi:10.1016/S0045-7825(99)00394-1
[27] Quagliarella, D, Iannelli, P, Vitagliano, PL and Chinnici, G. Aerodynamic Shape Design Using Hybrid Evolutionary Computation and Fitness Approximation. Proceedings of the 1stAIAA Intelligent Systems Technical Conference. 2004. pp.971–981. American Institute of Aeronautics and Astronautics.
[28] DOI: 10.1016/S0954-1810(96)00050-7 · Zbl 05388382 · doi:10.1016/S0954-1810(96)00050-7
[29] DOI: 10.1109/T-C.1969.222678 · doi:10.1109/T-C.1969.222678
[30] Schaback R, Approximation Theory VIII pp 491– (1995)
[31] Sefrioui, M. and Périaux, J. (2000), ’A Hierarchical Genetic Algorithm Using Multiple Models for Optimization’, inProceedings of the 6th International Conference on Parallel Problem Solving from Nature–PPSN VI, eds. M. Schoenauer, K. Deb, G. Rudolph, X. Yao, E. Lutton, J.J.M. Guervós, and H.P. Schwefel, no. 1917 in Lecture Notes in Computer Science, Springer-Verlag, pp. 879–888
[32] DOI: 10.1007/BF01197554 · doi:10.1007/BF01197554
[33] DOI: 10.1214/aos/1176345969 · Zbl 0511.62048 · doi:10.1214/aos/1176345969
[34] DOI: 10.1007/978-3-540-49774-5_17 · doi:10.1007/978-3-540-49774-5_17
[35] DOI: 10.1007/978-3-540-76286-7_3 · doi:10.1007/978-3-540-76286-7_3
[36] DOI: 10.1007/s00500-008-0348-2 · Zbl 05558132 · doi:10.1007/s00500-008-0348-2
[37] Tenne, Y. and Goh, C.K. (eds.) (2010),Computational Intelligence in Expensive Optimization Problems, Vol. 2 ofEvolutionary Learning and Optimization, Springer · Zbl 1187.90020
[38] Tenne, Y. Izui, K., and Nishiwaki, S. (2010a), ’Handling Undefined Vectors in Expensive Optimization Problems’, inProceedings of the 2010 EvoStar Conference, ed. C. Di Chio, Vol. 6024/2010 of Lecture Notes in Computer Science, Berlin: Springer, pp. 582–591
[39] Tenne, Y, Izui, K and Nishiwaki, S. Dimensionality Reduction Frameworks for Computationally Expensive Problems. Proceedings of the 2010 IEEE Conference on Evolutionary Computation–CEC 2010. 2010b. Piscataway, New Jersey: IEEE. · Zbl 1307.90208
[40] Vapnik VN, Statistical Learning Theory (1998)
This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. In some cases that data have been complemented/enhanced by data from zbMATH Open. This attempts to reflect the references listed in the original paper as accurately as possible without claiming completeness or a perfect matching.