💥1 概述
狼优化（GWO）算法是基于灰狼社会等级及其狩猎与合作策略的新兴算法。该算法于2014年推出，已被大量研究人员和设计师使用，因此原始论文的引用次数超过了许多其他算法。在Niu等人最近的一项研究中，介绍了这种算法优化现实世界问题的主要缺点之一。总之，他们表明，随着问题的最佳解决方案从0发散，GWO的性能会下降。
在贪婪非分层灰狼优化器（G-NHGWO）中，通过对原始GWO算法进行直接修改，即忽略其社会等级，我们能够在很大程度上消除这一缺陷，并为该算法的未来使用开辟了新的视角。通过将所提方法应用于基准和实际工程问题，验证了该方法的效率。
文章下载链接：
http://dx.doi.org/10.1049/ell2.12176
📚2 运行结果
部分代码：
clc clear global NFE NFE=0; nPop=30; % Number of search agents (Population Number) MaxIt=1000; % Maximum number of iterations nVar=30; % Number of Optimization Variables Total_Runs=25; % Pre-allocating vectors and matrices Cost_Rsult=nan(1,Total_Runs); Rsult=nan(Total_Runs,MaxIt); Mean=nan(1,14); Best=Mean; Std=Mean; nfe=Mean; fitness1=nan(1,nPop); for nFun=1:14 NFE=0; CostFunction=@(x,nFun) Cost(x,nFun); % Cost Function for run_no=1:Total_Runs %% Problem Definition VarMin=-100; % Decision Variables Lower Bound if nFun==7 VarMin=-600; % Decision Variables Lower Bound end if nFun==8 VarMin=-32; % Decision Variables Lower Bound end if nFun==9 VarMin=-5; % Decision Variables Lower Bound end if nFun==10 VarMin=-5; % Decision Variables Lower Bound end if nFun==11 VarMin=-0.5; % Decision Variables Lower Bound end if nFun==12 VarMin=-pi; % Decision Variables Lower Bound end if nFun==14 VarMin=-100; % Decision Variables Lower Bound end VarMax= -VarMin; % Decision Variables Upper Bound if nFun==13 VarMin=-3; % Decision Variables Lower Bound VarMax= 1; % Decision Variables Upper Bound end %% Grey Wold Optimizer (GWO) % Initialize Best Solution (Alpha) which will be used for archiving Alpha_pos=zeros(1,nVar); Alpha_score=inf; % Initialize the positions of search agents Positions=rand(nPop,nVar).*(VarMax-VarMin)+VarMin; Positions1=rand(nPop,nVar).*(VarMax-VarMin)+VarMin; BestCosts=zeros(1,MaxIt); fitness(1:nPop)=inf; iter=0; % Loop counter %% Main loop while iter<MaxIt for i=1:nPop % Return back the search agents that go beyond the boundaries of the search space Flag4ub=Positions1(i,:)>VarMax; Flag4lb=Positions1(i,:)<VarMin; Positions1(i,:)=(Positions1(i,:).*(~(Flag4ub+Flag4lb)))+VarMax.*Flag4ub+VarMin.*Flag4lb; % Calculate objective function for each search agent fitness1(i)= CostFunction(Positions1(i,:), nFun); % Grey Wolves if fitness1(i)<fitness(i) Positions(i,:)=Positions1(i,:); fitness(i) =fitness1(i) ; end % Update Best Solution (Alpha) for archiving if fitness(i)<Alpha_score Alpha_score=fitness(i); Alpha_pos=Positions(i,:); end end a=2-(iter*((2)/MaxIt)); % a decreases linearly fron 2 to 0 % Update the Position of all search agents for i=1:nPop for j=1:nVar GGG=randperm(nPop-1,3); ind1= GGG>=i; GGG(ind1)=GGG(ind1)+1; m1=GGG(1); m2=GGG(2); m3=GGG(3); r1=rand; r2=rand; A1=2*a*r1-a; C1=2*r2; D_alpha=abs(C1*Positions(m1,j)-Positions(i,j)); X1=Positions(m1,j)-A1*D_alpha; r1=rand; r2=rand; A2=2*a*r1-a; C2=2*r2; D_beta=abs(C2*Positions(m2,j)-Positions(i,j)); X2=Positions(m2,j)-A2*D_beta; r1=rand; r2=rand; A3=2*a*r1-a; C3=2*r2; D_delta=abs(C3*Positions(m3,j)-Positions(i,j)); X3=Positions(m3,j)-A3*D_delta; Positions1(i,j)=(X1+X2+X3)/3; end end iter=iter+1; BestCosts(iter)=Alpha_score; fprintf('Func No= %-2.0f, Run No= %-2.0f, Iter= %g, Best Cost = %g\n',nFun,run_no,iter,Alpha_score); end %%% Results Cost_Rsult(1, run_no)=Alpha_score; Rsult(run_no,:)= BestCosts; end Mean(nFun)=mean(Cost_Rsult); Best(nFun)=min(Cost_Rsult); Std(nFun)=std(Cost_Rsult); nfe(nFun)=NFE;
🌈3 Matlab代码实现
🎉4 参考文献
部分理论来源于网络，如有侵权请联系删除。
[1]Ebrahim Akbari (2022). Greedy Non‐Hierarchical Grey Wolf Optimizer (G-NHGWO)
[2]Akbari, Ebrahim, et al. “A Greedy Non-Hierarchical Grey Wolf Optimizer for Real-World Optimization.” Electronics Letters, Institution of Engineering and Technology (IET), Apr. 2021, doi:10.1049/ell2.12176.
