【智能优化算法】卷尾猴搜索算法(Capuchin search algorithm,CapSA)

【智能优化算法】卷尾猴搜索算法(Capuchin search algorithm,CapSA)是期刊“NEURAL COMPUTING & APPLICATIONS”（IF 6.0）的2021年智能优化算法

01.引言

【智能优化算法】卷尾猴搜索算法(Capuchin search algorithm,CapSA)用于解决约束和全局优化问题。CapSA的主要灵感来自卷尾猴的动态行为。该算法的基本优化特征是通过模拟卷尾猴在森林中寻找食物来源时在树木和河岸上徘徊和觅食的社会行为来设计的。卷尾猴在觅食过程中的一些常见行为是在这个算法中实现的，包括跳跃、摇摆和攀爬。跳跃是卷尾猴从一棵树跳到另一棵树的有效机制。卷尾猴的另一种觅食机制是摇摆和攀爬，它允许卷尾猴在树木、树枝和树枝的末端上移动一小段距离。这些运动机制最终导致全局优化问题的可行解。该算法在23个著名的基准函数上进行了基准测试，并解决了一些具有挑战性和计算成本高的工程问题。我们进行了一项广泛的比较研究，以证明CapSA在优化精度和统计检验分析方面优于几种突出的元启发式算法。总体结果表明，与竞争的元启发式方法相比，CapSA提供了更精确的解，具有更高的收敛速度。

02.优化算法的流程

03.论文中算法对比图

04.部分代码

function [fitCapSA,CapPos,cg_curve]=CapSA(noP,maxite,LB,UB,dim,fobj)warning off; format long;
%%%%* 1
% if size(UB,2)==1
%     UB=ones(1,dim)*UB;
%     LB=ones(1,dim)*LB;
% end% if size(UB,1)==1
%     UB=ones(dim,1)*UB;
%     LB=ones(dim,1)*LB;
% end
% % % CapSA main program
% f1 =  figure (1);
% set(gcf,'color','w');
% hold on
% xlabel('Iteration','interpreter','latex','FontName','Times','fontsize',10)
% ylabel('Fit value','interpreter','latex','FontName','Times','fontsize',10); 
% grid;
%  
cg_curve=zeros(1,maxite);
%% % CapSA initialization
%Initialize the Pos of Caps in the space
CapPos=initialization(noP,dim,UB,LB);
v=0.1*CapPos;% initial velocity
v0=zeros(noP,dim); 
CapFit=zeros(noP,1);
% Calculate the Fit of initialCapsfor i=1:noPCapFit(i,1)=fobj(CapPos(i,:));end% Initial Fit of the random Pos
Fit = CapFit;
[fitCapSA,index]=min(CapFit);
CapBestPos = CapPos; % Best Pos initialization
Pos= CapPos; 
gFoodPos = CapPos(index,:); % initial global Pos 
%% % CapSA Parameters
bf=0.70;%Balance factor
cr=11.0;  %Modulus of elasticity
g=9.81;
% CapSA velocity updates
a1=1.250; a2=1.5;   
beta=[2 11 2];
wmax=0.8;
wmin=0.1;%% % CapSA Main loopfor t = 1 : maxite% Life time convergencetau = beta(1) * exp(-beta(2) * t/maxite)^beta(3);w   = wmax-(wmax-wmin)*(t/maxite); fol=ceil((noP-1).*rand(noP,1))'; %  % CapSA velocity update
for i=1:noP for j=1:dimv(i,j)=  w* v(i,j) + ...a1*(CapBestPos(i,j)- CapPos(i,j))*rand + ...a2*(gFoodPos(j)  - CapPos(i,j))*rand;      end
end% CapSA Pos update
for i=1:noPif i<noP/2if (rand()>=0.1)r=rand;if r<=0.15CapPos(i,:) =  gFoodPos +    bf*((v(i,:).^2)*sin(2*rand()*1.5))/g;  % Jumping (Projection)elseif   r>0.15 && r<=0.30  CapPos(i,:) =  gFoodPos +   cr*bf*((v(i,:).^2)*sin(2*rand()*1.5))/g;  % Jumping (Land)  elseif   r>0.30 && r<=0.9      CapPos(i,:) =    CapPos(i,:) +  v(i,:); % Movement on the ground    elseif  r>0.9 && r<=0.95 CapPos(i,:) =      gFoodPos   +  bf*sin(rand()*1.5);   % Swing % Local search  elseif   r>0.95 CapPos(i,:) =       gFoodPos   +  bf*(v(i,:)- v0(i,:));    % Climbing   % Local searchendelseCapPos(i,:) =           tau*(LB  + rand *(UB- LB));   end% Let the followers follow the leaders (update their Pos)
elseif i>=noP/2 && i<=noP eps=((rand()+2*rand())-(3*rand()))/(1+rand()); Pos(i,:)=gFoodPos+2*(CapBestPos(fol(i),:)-CapPos(i,:))*eps +...2*(CapPos(i,:)-CapBestPos(i,:))*eps;      CapPos(i,:)=(Pos(i,:)+ CapPos(i-1,:))/(2); end
end 
v0 = v;   for i=1:noP % relocation (Update, exploration)u=UB-CapPos(i,:)<0;l=LB-CapPos(i,:)>0;CapPos(i,:)= LB.*l+UB.*u+CapPos(i,:).*~xor(u,l);CapFit(i,1)=fobj (CapPos(i,:)) ;if CapFit(i,1)<Fit(i)CapBestPos(i,:)=CapPos(i,:);Fit(i)=CapFit(i,1);end 
end
%% Evaluate the new Pos
[fmin,index]=min(Fit); % finding out the best Pos  
% Updating gPos and best Fit
if fmin < fitCapSAgFoodPos = CapBestPos(index,:); % Update the global best PosfitCapSA = fmin;
end
% %     % Display the iterative results
% 
%          disp(['Iteration# ', num2str(t) , '  Fit= ' , num2str(fitCapSA)]);
% %    % Obtain the convergence curvecg_curve(t)=fitCapSA;%  if t>2
%      set(0, 'CurrentFigure', f1)
% 
%         line([t-1 t], [cg_curve(t-1) cg_curve(t)],'Color','b'); 
%         xlabel('Iteration');
%         ylabel('Best score obtained so far');
%         drawnow 
%  end
%   end
fitCapSA =fobj(gFoodPos);
end