文章转自:http://www.matlabsky.com/thread-12414-1-1.html
基于GridSearch的svm参数寻优
http://www.matlabsky.com/thread-12411-1-1.html
基于GA的svm参数寻优
http://www.matlabsky.com/thread-12412-1-1.html
====================================================
在这里使用启发式算法PSO来进行参数寻优,用网格划分(grid search)来寻找最佳的参数c和g,虽然采用网格搜索能够找到在CV意义下的最高的分类准确率,即全局最优解,但有时候如果想在更大的范围内寻找最佳的参数c和g会很费时,采用启发式算法就可以不必遍历网格内的所有的参数点,也能找到全局最优解。
关于粒子群优化算法这里不打算过多介绍,想要学习的朋友可以自己查看相关资料。
使用PSO来进行参数寻优在在libsvm-mat-2.89-3[FarutoUltimate3.0]工具箱中已经实现psoSVMcgForClass.m(分类问题参数寻优)、psoSVMcgForRegress.m(回归问题参数寻优)。
函数使用接口:
- 利用PSO参数寻优函数(分类问题):psoSVMcgForClass
- [bestCVaccuracy,bestc,bestg,pso_option]=
- psoSVMcgForClass(train_label,train,pso_option)
- 输入:
- train_label:训练集的标签,格式要求与svmtrain相同。
- train:训练集,格式要求与svmtrain相同。
- pso_option:PSO中的一些参数设置,可不输入,有默认值,详细请看代码的帮助说明。
- 输出:
- bestCVaccuracy:最终CV意义下的最佳分类准确率。
- bestc:最佳的参数c。
- bestg:最佳的参数g。
- pso_option:记录PSO中的一些参数。
- ==========================================================
- 利用PSO参数寻优函数(回归问题):psoSVMcgForRegress
- [bestCVmse,bestc,bestg,pso_option]=
- psoSVMcgForRegress(train_label,train,pso_option)
- 其输入输出与psoSVMcgForClass类似,这里不再赘述。
- function [bestCVaccuarcy,bestc,bestg,pso_option] = psoSVMcgForClass(train_label,train,pso_option)
- % psoSVMcgForClass
-
- %%
- % by faruto
- %Email:patrick.lee@foxmail.com QQ:516667408 http://blog.sina.com.cn/faruto BNU
- %last modified 2010.01.17
-
- %% 若转载请注明:
- % faruto and liyang , LIBSVM-farutoUltimateVersion
- % a toolbox with implements for support vector machines based on libsvm, 2009.
- %
- % Chih-Chung Chang and Chih-Jen Lin, LIBSVM : a library for
- % support vector machines, 2001. Software available at
- % http://www.csie.ntu.edu.tw/~cjlin/libsvm
- %% 参数初始化
- if nargin == 2
- pso_option = struct('c1',1.5,'c2',1.7,'maxgen',200,'sizepop',20, ...
- 'k',0.6,'wV',1,'wP',1,'v',5, ...
- 'popcmax',10^2,'popcmin',10^(-1),'popgmax',10^3,'popgmin',10^(-2));
- end
- % c1:初始为1.5,pso参数局部搜索能力
- % c2:初始为1.7,pso参数全局搜索能力
- % maxgen:初始为200,最大进化数量
- % sizepop:初始为20,种群最大数量
- % k:初始为0.6(k belongs to [0.1,1.0]),速率和x的关系(V = kX)
- % wV:初始为1(wV best belongs to [0.8,1.2]),速率更新公式中速度前面的弹性系数
- % wP:初始为1,种群更新公式中速度前面的弹性系数
- % v:初始为3,SVM Cross Validation参数
- % popcmax:初始为100,SVM 参数c的变化的最大值.
- % popcmin:初始为0.1,SVM 参数c的变化的最小值.
- % popgmax:初始为1000,SVM 参数g的变化的最大值.
- % popgmin:初始为0.01,SVM 参数c的变化的最小值.
-
- Vcmax = pso_option.k*pso_option.popcmax;
- Vcmin = -Vcmax ;
- Vgmax = pso_option.k*pso_option.popgmax;
- Vgmin = -Vgmax ;
-
- eps = 10^(-3);
-
- %% 产生初始粒子和速度
- for i=1:pso_option.sizepop
-
- % 随机产生种群和速度
- pop(i,1) = (pso_option.popcmax-pso_option.popcmin)*rand+pso_option.popcmin;
- pop(i,2) = (pso_option.popgmax-pso_option.popgmin)*rand+pso_option.popgmin;
- V(i,1)=Vcmax*rands(1,1);
- V(i,2)=Vgmax*rands(1,1);
-
- % 计算初始适应度
- cmd = ['-v ',num2str(pso_option.v),' -c ',num2str( pop(i,1) ),' -g ',num2str( pop(i,2) )];
- fitness(i) = svmtrain(train_label, train, cmd);
- fitness(i) = -fitness(i);
- end
-
- % 找极值和极值点
- [global_fitness bestindex]=min(fitness); % 全局极值
- local_fitness=fitness; % 个体极值初始化
-
- global_x=pop(bestindex,:); % 全局极值点
- local_x=pop; % 个体极值点初始化
-
- % 每一代种群的平均适应度
- avgfitness_gen = zeros(1,pso_option.maxgen);
-
- %% 迭代寻优
- for i=1:pso_option.maxgen
-
- for j=1:pso_option.sizepop
-
- %速度更新
- V(j,:) = pso_option.wV*V(j,:) + pso_option.c1*rand*(local_x(j,:) - pop(j,:)) + pso_option.c2*rand*(global_x - pop(j,:));
- if V(j,1) > Vcmax
- V(j,1) = Vcmax;
- end
- if V(j,1) < Vcmin
- V(j,1) = Vcmin;
- end
- if V(j,2) > Vgmax
- V(j,2) = Vgmax;
- end
- if V(j,2) < Vgmin
- V(j,2) = Vgmin;
- end
-
- %种群更新
- pop(j,:)=pop(j,:) + pso_option.wP*V(j,:);
- if pop(j,1) > pso_option.popcmax
- pop(j,1) = pso_option.popcmax;
- end
- if pop(j,1) < pso_option.popcmin
- pop(j,1) = pso_option.popcmin;
- end
- if pop(j,2) > pso_option.popgmax
- pop(j,2) = pso_option.popgmax;
- end
- if pop(j,2) < pso_option.popgmin
- pop(j,2) = pso_option.popgmin;
- end
-
- % 自适应粒子变异
- if rand>0.5
- k=ceil(2*rand);
- if k == 1
- pop(j,k) = (20-1)*rand+1;
- end
- if k == 2
- pop(j,k) = (pso_option.popgmax-pso_option.popgmin)*rand + pso_option.popgmin;
- end
- end
-
- %适应度值
- cmd = ['-v ',num2str(pso_option.v),' -c ',num2str( pop(j,1) ),' -g ',num2str( pop(j,2) )];
- fitness(j) = svmtrain(train_label, train, cmd);
- fitness(j) = -fitness(j);
-
- cmd_temp = ['-c ',num2str( pop(j,1) ),' -g ',num2str( pop(j,2) )];
- model = svmtrain(train_label, train, cmd_temp);
-
- if fitness(j) >= -65
- continue;
- end
-
- %个体最优更新
- if fitness(j) < local_fitness(j)
- local_x(j,:) = pop(j,:);
- local_fitness(j) = fitness(j);
- end
-
- if abs( fitness(j)-local_fitness(j) )<=eps && pop(j,1) < local_x(j,1)
- local_x(j,:) = pop(j,:);
- local_fitness(j) = fitness(j);
- end
-
- %群体最优更新
- if fitness(j) < global_fitness
- global_x = pop(j,:);
- global_fitness = fitness(j);
- end
-
- if abs( fitness(j)-global_fitness )<=eps && pop(j,1) < global_x(1)
- global_x = pop(j,:);
- global_fitness = fitness(j);
- end
-
- end
-
- fit_gen(i) = global_fitness;
- avgfitness_gen(i) = sum(fitness)/pso_option.sizepop;
- end
-
- %% 结果分析
- figure;
- hold on;
- plot(-fit_gen,'r*-','LineWidth',1.5);
- plot(-avgfitness_gen,'o-','LineWidth',1.5);
- legend('最佳适应度','平均适应度',3);
- xlabel('进化代数','FontSize',12);
- ylabel('适应度','FontSize',12);
- grid on;
-
- % print -dtiff -r600 pso
-
- bestc = global_x(1);
- bestg = global_x(2);
- bestCVaccuarcy = -fit_gen(pso_option.maxgen);
-
- line1 = '适应度曲线Accuracy[PSOmethod]';
- line2 = ['(参数c1=',num2str(pso_option.c1), ...
- ',c2=',num2str(pso_option.c2),',终止代数=', ...
- num2str(pso_option.maxgen),',种群数量pop=', ...
- num2str(pso_option.sizepop),')'];
- % line3 = ['Best c=',num2str(bestc),' g=',num2str(bestg), ...
- % ' CVAccuracy=',num2str(bestCVaccuarcy),'%'];
- % title({line1;line2;line3},'FontSize',12);
- title({line1;line2},'FontSize',12);