精英教學(xué)優(yōu)化算法（Elitist teaching-learning-based optimization,ETLBO）是一種基于實際班級教學(xué)過程的新型優(yōu)化算法。針對ETLBO算法存在的尋優(yōu)精度低、穩(wěn)定性差的問題,提出一種基于自適應(yīng)反饋機制的精英教學(xué)優(yōu)化算法（Adaptive Feedback ETLBO,AFETLBO）。在學(xué)生階段之后,通過添加自適應(yīng)反饋機制,將學(xué)生分為優(yōu)等生和差生,且動態(tài)調(diào)整兩者的規(guī)模,對差生實行與教師之間的反饋交流,快速向教師靠攏,加強收斂能力;對優(yōu)等生實行自我學(xué)習(xí),進行局部精細(xì)搜索。自適應(yīng)反饋階段的加入,增加了學(xué)習(xí)方式,保持了學(xué)生的多樣性特性,提高全局搜索能力。對6個無約束及5個標(biāo)準(zhǔn)函數(shù)的測試結(jié)果表明,與其他優(yōu)化算法相比,AFETLBO算法具有更高的尋優(yōu)精度和收斂能力。 

【文章來源】：系統(tǒng)仿真學(xué)報. 2018,30(08)北大核心

【文章頁數(shù)】：8 頁

【部分圖文】：

各算法在f1的表現(xiàn)Fig.2Performanceofalgorithmsinf1

第30卷第8期Vol.30No.82018年8月李榮雨,等:基于自適應(yīng)反饋機制的精英教學(xué)優(yōu)化算法Aug.,2018http:∥www.china-simulation.com2955表2無約束測試函數(shù)對比結(jié)果平均值(標(biāo)準(zhǔn)差)Tab.2Comparisonofresultsforunconstrainedbenchmarkfunctionsmean(std)函數(shù)維數(shù)TLBOETLBOFETLBOAFETLBOf1309.86E–13(7.36E–15)3.97E–165(2.54E–164)3.43E–231(3.21E–231)0.00E+000(0.0E+000)1002.31E–10(9.43E–12)1.50E–163(2.26E–163)2.61E–230(1.12E–230)0.00E+000(0.0E+000)f2306.62E–07(2.35E–08)6.21E–015(1.87E–015)4.44E–015(0.0E+00)1.01E–017(4.23E–016)1009.26E–08(1.01E–07)6.21E–015(1.99E–014)4.44E–015(0.0E+00)1.52E–017(1.68E–017)f3301.52E–16(5.76E–15)0.0E+000(0.0E+000)0.0E+000(0.0E+000)0.0E+000(0.0E+000)1003.64E–12(4.36E–12)0.0E+000(0.0E+000)0.0E+000(0.0E+000)0.0E+000(0.0E+000)f4305.87E+01(1.68E+1)2.67E+001(2.26E–01)2.57E+001(1.23E–01)1.05E+000(6.32E–001)1007.35E+00(1.98E+0)2.67E+001(3.31E–01)2.60E+001(2.16E–01)5.16E+000(1.3E+000)f5304.67E–25(5.12E–24)0.0E+000(0.0E–000)0.0E+000(0.0E–000)0.0E+000(0.0E–000)1008.63E–23(1.93E–24)0.0E+000(0.0E–000)0.0E+000(0.0E–000)0.0E+000(0.0E–000)f6307.27E–07(6.61E–08)1.26E–083(1.76E–083)4.16E–116(1.07E–115)1.09E–139(1.01E–139)1008.91E–07(1.99E–7)2.73E–083(3.96E–083)4.49E–115(8.74E–115)2.71E–138(1.56E–137)圖2各算法在f1的表現(xiàn)圖3各算法在f2的表現(xiàn)Fig.2Performanceofalgorithmsinf1Fig.3Performanceofalgorithmsinf2圖4各算法在f3的

第30卷第8期Vol.30No.82018年8月李榮雨,等:基于自適應(yīng)反饋機制的精英教學(xué)優(yōu)化算法Aug.,2018http:∥www.china-simulation.com2955表2無約束測試函數(shù)對比結(jié)果平均值(標(biāo)準(zhǔn)差)Tab.2Comparisonofresultsforunconstrainedbenchmarkfunctionsmean(std)函數(shù)維數(shù)TLBOETLBOFETLBOAFETLBOf1309.86E–13(7.36E–15)3.97E–165(2.54E–164)3.43E–231(3.21E–231)0.00E+000(0.0E+000)1002.31E–10(9.43E–12)1.50E–163(2.26E–163)2.61E–230(1.12E–230)0.00E+000(0.0E+000)f2306.62E–07(2.35E–08)6.21E–015(1.87E–015)4.44E–015(0.0E+00)1.01E–017(4.23E–016)1009.26E–08(1.01E–07)6.21E–015(1.99E–014)4.44E–015(0.0E+00)1.52E–017(1.68E–017)f3301.52E–16(5.76E–15)0.0E+000(0.0E+000)0.0E+000(0.0E+000)0.0E+000(0.0E+000)1003.64E–12(4.36E–12)0.0E+000(0.0E+000)0.0E+000(0.0E+000)0.0E+000(0.0E+000)f4305.87E+01(1.68E+1)2.67E+001(2.26E–01)2.57E+001(1.23E–01)1.05E+000(6.32E–001)1007.35E+00(1.98E+0)2.67E+001(3.31E–01)2.60E+001(2.16E–01)5.16E+000(1.3E+000)f5304.67E–25(5.12E–24)0.0E+000(0.0E–000)0.0E+000(0.0E–000)0.0E+000(0.0E–000)1008.63E–23(1.93E–24)0.0E+000(0.0E–000)0.0E+000(0.0E–000)0.0E+000(0.0E–000)f6307.27E–07(6.61E–08)1.26E–083(1.76E–083)4.16E–116(1.07E–115)1.09E–139(1.01E–139)1008.91E–07(1.99E–7)2.73E–083(3.96E–083)4.49E–115(8.74E–115)2.71E–138(1.56E–137)圖2各算法在f1的表現(xiàn)圖3各算法在f2的表現(xiàn)Fig.2Performanceofalgorithmsinf1Fig.3Performanceofalgorithmsinf2圖4各算法在f3的

【參考文獻(xiàn)】：
期刊論文
[1]基于混合策略的自適應(yīng)教與學(xué)優(yōu)化算法[J]. 畢曉君,李月,陳春雨.  哈爾濱工程大學(xué)學(xué)報. 2016(06)
[2]多學(xué)習(xí)教與學(xué)優(yōu)化算法[J]. 李志南,南新元,李娜,史德生.  計算機應(yīng)用與軟件. 2016(02)
[3]一種用于PID控制的教與學(xué)優(yōu)化算法[J]. 拓守恒,雍龍泉.  智能系統(tǒng)學(xué)報. 2014(06)
[4]基于反饋的精英教學(xué)優(yōu)化算法[J]. 于坤杰,王昕,王振雷.  自動化學(xué)報. 2014(09)
[5]基于差分進化與群搜索的混合優(yōu)化算法及在乙烯裂解爐中的應(yīng)用(英文)[J]. 年笑宇,王振雷,錢鋒.  Chinese Journal of Chemical Engineering. 2013(05)



本文編號：3046339