【智能优化算法】金枪鱼群优化（Tuna Swarm Optimization,TSO）

金枪鱼群优化（Tuna Swarm Optimization,TSO）是期刊“Computational Intelligence and Neuroscience”（IF：1.8）的2021年智能优化算法

01.引言

金枪鱼群优化（Tuna Swarm Optimization,TSO）的主要灵感来自于金枪鱼群体的合作觅食行为。模拟金枪鱼群体的两种觅食行为，包括螺旋觅食和抛物线觅食，以开发一种有效的元启发式算法。在一组基准函数和几个真实的工程问题上，通过与其他元算法的比较，对TSO的性能进行了评估。使用灵敏度、可扩展性、稳健性和收敛性分析，并结合Wilcoxon秩和检验和Friedman检验。仿真结果表明，TSO算法的性能优于其他比较算法。

02.优化算法的流程

03.论文中算法对比图

04.部分代码

function [Tuna1_fit,Tuna1,Convergence_curve]=TSO(Particles_no,Max_iter,Low,Up,Dim,fobj)
Tuna1=zeros(1,Dim);   Tuna1_fit=inf;
T=initialization(Particles_no,Dim,Up,Low);
Iter=0;
aa=0.7;
z=0.05;
while Iter<Max_iter
    C=Iter/Max_iter;
    a1=aa+(1-aa)*C;
    a2=(1-aa)-(1-aa)*C;
    for i=1:size(T,1)
        
        Flag4ub=T(i,:)>Up;
        Flag4lb=T(i,:)<Low;
        T(i,:)=(T(i,:).*(~(Flag4ub+Flag4lb)))+Up.*Flag4ub+Low.*Flag4lb;
        
        fitness(i)=fobj(T(i,:));
        
        if fitness(i)<Tuna1_fit
            Tuna1_fit=fitness(i);  Tuna1=T(i,:);
        end
    end
    
    %---------------- Memory saving-------------------
    if Iter==0
        fit_old=fitness;  C_old=T;
    end
    
    for i=1:Particles_no
        if fit_old(i)<fitness(i)
            fitness(i)=fit_old(i); T(i,:)=C_old(i,:);
        end
    end
    
    C_old=T;  fit_old=fitness;
    %-------------------------------------------------
    
    t=(1-Iter/Max_iter)^(Iter/Max_iter);                   
    
    if rand<z
        T(1,:)= (Up-Low)*rand+Low;
    else
        if  0.5<rand
            r1=rand;
            Beta=exp(r1*exp(3*cos(pi*((Max_iter-Iter+1)/Max_iter))))*(cos(2*pi*r1));
            if  C>rand
                T(1,:)=a1.*(Tuna1+Beta*abs(Tuna1-T(1,:)))+a2.*T(1,:); %Equation (8.3)
                
            else
                IndivRand=rand(1,Dim).*(Up-Low)+Low;
                T(1,:)=a1.*(IndivRand+Beta*abs(IndivRand-T(i,:)))+a2.*T(1,:);%Equation (8.1)
            end
        else
            TF = (rand>0.5)*2-1;
            if 0.5>rand
                T(1,:)=Tuna1+rand(1,Dim).*(Tuna1-T(1,:))+TF.*t^2.*(Tuna1-T(1,:));%Equation (9.1)
            else
                T(1,:) =TF.* t^2.*T(1,:);%Equation (9.2)
            end
            
        end
        
    end
    
    for i=2:Particles_no
        if rand<z    
            
            T(i,:)= (Up-Low)*rand+Low;
        else
            if  0.5<rand
                r1=rand;
                Beta=exp(r1*exp(3*cos(pi*((Max_iter-Iter+1)/Max_iter))))*(cos(2*pi*r1));
                if  C>rand
                    T(i,:)=a1.*(Tuna1+Beta*abs(Tuna1-T(i,:)))+a2.*T(i-1,:);%Equation (8.4)
                else
                    
                    IndivRand=rand(1,Dim).*(Up-Low)+Low;
                    T(i,:)=a1.*(IndivRand+Beta*abs(IndivRand-T(i,:)))+a2.*T(i-1,:);%Equation (8.2)
                end
            else
                TF = (rand>0.5)*2-1;
                if 0.5>rand
                    T(i,:)=Tuna1+rand(1,Dim).*(Tuna1-T(i,:))+TF*t^2.*(Tuna1-T(i,:)); %Equation (9.1)
                else
                    T(i,:) = TF*t^2.*T(i,:);%Equation (9.2)
                end
            end
        end
    end
    
    Iter=Iter+1;
    Convergence_curve(Iter)=Tuna1_fit;
    
end