遗传算法matlab程序实例.doc

--------------------------------------------------------------------------------------------------遗传算法程序(一)说明 fga.m 为遗传算法的主程序 ; 采用二进制 Gray 编码 ,采用基于轮盘赌法的非线性排名选择, 均匀交叉 ,变异操作,而且还引入了倒位操作function [BestPop,Trace]fgaFUN,LB,UB,eranum,popsize,pCross,pMutation,pInversion,options [BestPop,Trace]fmaxgaFUN,LB,UB,eranum,popsize,pcross,pmutation Finds a maximum of a function of several variables. fmaxga solves problems of the max FX subject to LB 0error数据输入错误 ,请重新输入 LB1 时,bimodai-1ai,2其中原二进制串a1a2.an,Gray 串b1b2.bninitpopi,pop1end-1;endinitpoppopsize,ones1,len;The whole one encoding individual解码function [fval] b2fbval,bounds,bits fval - 表征各变量的十进制数 bval - 表征各变量的二进制编码串 bounds - 各变量的取值范围 bits - 各变量的二进制编码长度scalebounds,2-bounds,1./2.bits-1; The range of the variablesnumVsizebounds,1;cs[0 cumsumbits]; for i1numVabvalcsi1csi1;fvalisum2.sizea,2-1-10.*a*scaleiboundsi,1;end选择操作采用基于轮盘赌法的非线性排名选择各个体成员按适应值从大到小分配选择概率Piq/1-1-qn*1-qi, 其中 P0P1.Pn, sumPi1function [selectpop]NonlinearRankSelectFUN,pop,bounds,bits global m nselectpopzerosm,n;fitzerosm,1;for i1mfitifFUN1,,b2fpopi,,bounds,bits;以函数值为适应值做排名依据endselectprobfit/sumfit;计算各个体相对适应度0,1qmaxselectprob;选择最优的概率xzerosm,2;x,1[m-11];[y x,2]sortselectprob;rq/1-1-qm;标准分布基值newfitx,2r*1-q.x,1-1;生成选择概率newfitcumsumnewfit;计算各选择概率之和rNumssortrandm,1;fitIn1;newIn1;while newInpCross;lenlengthy1;if len2y1len[];endif lengthy12for i02lengthy1-2if opts0[NewPopy1i1,,NewPopy1i2,]EqualCrossOverOldPopy1i1,,OldPopy1i2,;else[NewPopy1i1,,NewPopy1i2,]MultiPointCrossOldPopy1i1,,OldPopy1i2,;endend endNewPopy2,OldPopy2,;采用均匀交叉function [children1,children2]EqualCrossOverparent1,parent2global n children1 children2 hidecoderoundrand1,n;随机生成掩码crosspositionfindhidecode1;holdpositionfindhidecode0;children1crosspositionparent1crossposition;掩码为 1，父 1 为子 1 提供基因children1holdpositionparent2holdposition;掩码为 0，父 2 为子 1 提供基因children2crosspositionparent2crossposition;掩码为 1，父 2 为子 2 提供基因children2holdpositionparent1holdposition;掩码为 0，父 1 为子 2 提供基因采用多点交叉，交叉点数由变量数决定function [Children1,Children2]MultiPointCrossParent1,Parent2global n Children1 Children2 VarNumChildren1Parent1;Children2Parent2;Pointssortunidrndn,1,2*VarNum;for i1VarNumChildren1Points2*i-1Points2*iParent2Points2*i-1Points2*i;Children2Points2*i-1Points2*iParent1Points2*i-1Points2*i;end变异操作function [NewPop]MutationOldPop,pMutation,VarNumglobal m n NewPoprrand1,m;positionfindr1for i1lenkunidrndn,1,VarNum; 设置变异点数，一般设置 1 点for j1lengthkif OldPoppositioni,kj1OldPoppositioni,kj0;elseOldPoppositioni,kj1;endendendendNewPopOldPop;倒位操作function [NewPop]InversionOldPop,pInversionglobal m n NewPopNewPopOldPop;rrand1,m;PopInfindr1for i1lendsortunidrndn,1,2;if d11NewPopPopIni,d1d2OldPopPopIni,d2-1d1;NewPopPopIni,d21nOldPopPopIni,d21n;endendend遗传算法程序(二)function youhuafunDcode; N50; Tunable maxgen50; Tunable crossrate0.5; Tunable muterate0.08; Tunable generation1; num lengthD; fatherrandrandintnum,N,3; score zerosmaxgen,N; while generation1450||minF2best.max_fadvance_kadvance_k1;x_betteradvance_kfitness.x;best.max_fmaxf;best.popuspopus;best.xfitness.x;endif modadvance_k,SPEEDUP_INTER0RANGEminmaxx_better;RANGEadvance0;end endreturn;function popusinit初始化M50;种群个体数目N30;编码长度popusroundrandM,N;return;function fitnessfitpopus,RANGE求适应度[M,N]sizepopus;fitnesszerosM,1;适应度fzerosM,1;函数值ARANGE1;BRANGE2;初始取值范围[0 255]for m1Mx0;for n1Nxxpopusm,n*2n-1;endxx*B-A/2NA;for k15fm,1fm,1-k*sink1*xk;endendf_stdf-minf./maxf-minf;函数值标准化fitness.ff;fitness.f_stdf_std;fitness.xx;return;function pickedchoosepopus,fitness选择ffitness.f;f_stdfitness.f_std;[M,N]sizepopus;choose_N3; 选择 choose_N 对双亲pickedzeroschoose_N,2; 记录选择好的双亲pzerosM,1; 选择概率d_orderzerosM,1;把父代个体按适应度从大到小排序f_tsortf,descend;将适应度按降序排列for k1Mxfindff_tk;降序排列的个体序号d_orderkx1;endfor m1Mpopus_tm,popusd_orderm,;endpopuspopus_t;ff_t;pf_std./sumf_std; 选择概率c_pcumsump; 累积概率for cn1choose_Npickedcn,1roulettec_p; 轮盘赌pickedcn,2roulettec_p; 轮盘赌popusintercrosspopus,pickedcn,;杂交endpopusaberrancepopus,picked;变异return;function popusintercrosspopus,picked 杂交[M_p,N_p]sizepicked;[M,N]sizepopus;for cn1M_pp1ceilrand*N;生成杂交位置p2ceilrand*N;psortp;tpopuspickedcn,1,p1p2;popuspickedcn,1,p1p2popuspickedcn,2,p1p2;popuspickedcn,2,p1p2t;endreturn;function popusaberrancepopus,picked 变异P_a0.05;变异概率[M,N]sizepopus;[M_p,N_p]sizepicked;Urand1,2;for kp1M_pif U2P_a 如果大于变异概率，就不变异continue;endif U10.5apickedkp,1;elseapickedkp,2;endp1ceilrand*N;生成变异位置p2ceilrand*N;if popusa,p110 1 变换popusa,p10;elsepopusa,p11;endif popusa,p21popusa,p20;elsepopusa,p21;endendreturn;function pickedroulettec_p 轮盘赌[M,N]sizec_p;Mmax[M N];Urand;if Uc_pm break;endend全方位的两点杂交、两点变异的改进的加速遗传算法(IAGA)