12.07.2015 Views

A Simulated Annealing Based Multi-objective Optimization Algorithm ...

A Simulated Annealing Based Multi-objective Optimization Algorithm ...

A Simulated Annealing Based Multi-objective Optimization Algorithm ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

91616161414141212121010108886664442220−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 10−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 10−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1(a) (b) (c)Fig. 6.The final non-dominated front obtained by (a) AMOSA (b) PAES (c) NSGA-II for SCH211.2110.80.50.60.50.40.2000−0.2−0.5−0.40 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9−0.60 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9(a) (b) (c)−0.50 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9Fig. 7.The final non-dominated front for Deb4 obtained by (a) AMOSA (b) PAES (c) NSGA-IITABLE IConvergence AND Purity MEASURES ON THE TEST FUNCTIONS FOR BINARY ENCODINGTest Convergence P urityProblem AMOSA PAES NSGA-II AMOSA PAES NSGA-IISCH1 0.0016 0.0016 0.0016 0.9950(99.5/100) 0.9850(98.5/100) 1(94/94)SCH2 0.0031 0.0015 0.0025 0.9950(99.5/100) 0.9670(96.7/100) 0.9974(97/97.3)ZDT1 0.0019 0.0025 0.0046 0.8350(83.5/100) 0.6535(65.4/100) 0.970(68.64/70.6)ZDT2 0.0028 0.0048 0.0390 0.8845(88.5/100) 0.4050(38.5/94.9) 0.7421(56.4/76)ZDT6 0.0026 0.0053 0.0036 1(100/100) 0.9949(98.8/99.3) 0.9880(66.5/67.3)Deb1 0.0046 0.0539 0.0432 0.91(91/100) 0.718(71.8/100) 0.77(71/92)Deb4 0.0026 0.0025 0.0022 0.98(98/100) 0.9522(95.2/100) 0.985(88.7/90)TABLE IISpacing AND MinimalSpacing MEASURES ON THE TEST FUNCTIONS FOR BINARY ENCODINGTest Spacing MinimalSpacingProblem AMOSA PAES NSGA-II AMOSA PAES NSGA-IISCH1 0.0167 0.0519 0.0235 0.0078 0.0530 0.0125SCH2 0.0239 0.5289 0.0495 N.A. N.A. N.A.ZDT 1 0.0097 0.0264 0.0084 0.0156 0.0265 0.0147ZDT 2 0.0083 0.0205 0.0079 0.0151 0.0370 0.0130ZDT 6 0.0051 0.0399 0.0089 0.0130 0.0340 0.0162Deb1 0.0166 0.0848 0.0475 0.0159 0.0424 0.0116Deb4 0.0053 0.0253 0.0089 N.A. N.A. N.A.is kept unlimited as in [17]. The results are compared tothose obtained by MOSA [17] and provided in [32].) AMOSAis executed for a total of 5000, 1000, 15000, 5000, 1000,5000 and 9000 run lengths respectively for DTLZ1, DTLZ2,DTLZ3, DTLZ4, DTLZ5, DTLZ5 and DTLZ6. Total numberof iterations, iter, per temperature is set accordingly. We haverun real-coded NSGA-II (code obtained from KANGAL site:http://www.iitk.ac.in/kangal/codes.html). For NSGA-II the followingparameter setting is used: probability of crossover=0.99, probability of mutation=(1/l), where l is the string

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!