main_par_kqaAnlysis.m

% 当前目录应为"标记分布工具包"
clear;

% 数据集字典
dataset = {'SBU_3DFE','SJAFFE','Yeast_spo5','Yeast_spo','Yeast_heat','Yeast_elu','Yeast_dtt','Yeast_diau','Yeast_cold','Yeast_cdc','Yeast_alpha','Flickr','Twitter','Human_Gene','RAF_ML','Natural_Scene'};

% 指定评估指标名称（若修改需同时修改ldlEvaluating函数）
indicatorName = {'Acc','LaLoss','KlDistance','EuclideanDistance','MSE','Chebyshev','Clark','Canberra','Cosine','Intersection','sortLoss','kurtosisKl','KurtLoss','SignedKurtosisOffset','AbsKurtosisOffset'};

% 实验所用算法
    lambda1 = [5e-1,1e-1,5e-2,1e-2,5e-3,1e-3,5e-4,1e-4,5e-5,1e-5,5e-6,1e-6,0];
    str = compose('%1.0e',lambda1);
    str = replace(str,'-','_');
    str = replace(str,'+','_');
    for i = 1:length(lambda1)
        algorithmName(i) = strcat('kqa_',str(i),'_');
    end
% 指定默认算法参数
    lambda2 = 1e-5;
    for i = 1:length(dataset)
        parms{i}.lambda1 = 1e-3;
        parms{i}.lambda2 = lambda2;
        parms{i}.method = 0;
        parms{i}.maxIter = 400;
        parms{i}.bfgs_lambda1 = parms{i}.lambda2;
    end
    parms{2}.maxIter = 250;
    parms{12}.maxIter = 200;
    parms{13}.maxIter = 200;
    parms{14}.maxIter = 200;

for datasetNum = 12:13  % 指定本次实验数据集编号范围
    
% 读取数据集并分为十折，返回值为size = (10,1)的元胞数组；
    datasetName = dataset{datasetNum};
    load( "dataSet\"+ datasetName+".mat");
    % isVaild——是否划分验证集; isRng——是否指定伪随机发生器
    nFold = 10; isVaild = false; isRng = true; 
    [trainFeatures,trainLabels,testFeatures,testLabels] = crossValidation(features,labels,nFold,isVaild,isRng);
    
% 生成算法名字符串常量，用于后面的结果评估表格
    algorithmName2 = cell(length(algorithmName),1);
    algorithmName3 = cell(2*length(algorithmName),1);
    for i = 1:length(algorithmName)
        algorithmName2{i} = strcat(upper(algorithmName{i}(1)),algorithmName{i}(2:end)); %#ok<*SAGROW>
        algorithmName3{2*i-1} = strcat(algorithmName{i},'Train');
        algorithmName3{2*i} = strcat(algorithmName{i},'Test');
    end

% 生成存储实验结果的表格
    for i = 1:length(algorithmName)
        eval([algorithmName{i},'Test = table;']);
        eval([algorithmName{i},'Train = table;']);
    end
    
% 模型训练及预测，返回评估指标（元胞数组）
    isParfor = true;    % isParfor——是否并行，非并行时部分算法会绘制loss曲线
%     for i = 1:length(algorithmName)
%         % try
%             eval(['[',algorithmName{i},'Test,',algorithmName{i},'Train] = parLdl',algorithmName2{i},'(trainFeatures,trainLabels,testFeatures,testLabels,parms{datasetNum},nFold,isParfor);']);
%         % catch ME
%         %    warning(getReport(ME));
%         %    continue;
%         % end
%     end

%     [~,~,bfgsModels] = parLdlBfgs(trainFeatures,trainLabels,testFeatures,testLabels,parms{datasetNum},nFold,isParfor);
%     parms{datasetNum}.models = bfgsModels;
    load("Ablation_KQA\lambda1_0.0001_lambda2_0\"+datasetName+".mat", "models");
    parms{datasetNum}.models = models.bfgsModels;
    
    for i = 1:length(lambda1)
        parms{datasetNum}.lambda1 = lambda1(i); 
        eval(['[',algorithmName{i},'Test,',algorithmName{i},'Train,',algorithmName{i},'Models] = parLdlKqa(trainFeatures,trainLabels,testFeatures,testLabels,parms{datasetNum},nFold,isParfor);']);
    end
   
    % 计算指标的均值和方差
    meanTest=[];
    meanTrain=[];
    meanAll=[];
    stdTest=[];
    stdTrain=[];
    stdAll=[];
    for i = 1:length(algorithmName)
        eval(['mean',algorithmName2{i},'Test = mean(',algorithmName{i},'Test{:,:},1);']);
        eval(['mean',algorithmName2{i},'Train = mean(',algorithmName{i},'Train{:,:},1);']);
        eval(['std',algorithmName2{i},'Test = std(',algorithmName{i},'Test{:,:},1);']);
        eval(['std',algorithmName2{i},'Train = std(',algorithmName{i},'Train{:,:},1);']);
        eval(['models.',algorithmName{i},'Models =', algorithmName{i},'Models;']);
        eval(['meanTest =[meanTest;mean',algorithmName2{i},'Test];']);
        eval(['meanTrain = [meanTrain;mean',algorithmName2{i},'Train];']);
        eval(['meanAll = [meanAll;mean',algorithmName2{i},'Train;mean',algorithmName2{i},'Test];']);
        eval(['stdTest =[stdTest;std',algorithmName2{i},'Test];']);
        eval(['stdTrain = [stdTrain;std',algorithmName2{i},'Train];']);
        eval(['stdAll = [stdAll;std',algorithmName2{i},'Train;std',algorithmName2{i},'Test];']);        
    end
    
    % 生成表格
    compareMeanTest = array2table(meanTest,'RowNames',algorithmName,'VariableNames',indicatorName);
    compareStdTest = array2table(stdTest,'RowNames',algorithmName,'VariableNames',indicatorName);
    compareMeanTrain = array2table(meanTrain,'RowNames',algorithmName,'VariableNames',indicatorName);
    compareStdTrain = array2table(stdTrain,'RowNames',algorithmName,'VariableNames',indicatorName);
    compareMeanAll = array2table(meanAll,'RowNames',algorithmName3,'VariableNames',indicatorName);
    compareStdAll = array2table(stdAll,'RowNames',algorithmName3,'VariableNames',indicatorName);

    % 保存结果，当前目录应为"标记分布工具包"
    cd('DataResult\ParamAnalysis_ECML');
    PARMs = parms{datasetNum};
    eval(['save ',datasetName,'.mat datasetName compareMeanAll compareMeanTest compareMeanTrain compareStdAll compareStdTest compareStdTrain PARMs models']);
    cd('..\..');

    clear stdKnnTest stdKnnTrain stdBfgsTest stdBfgsTrain stdLcTest stdLcTrain stdAdaboostBfgsTest stdAdaboostLcTest  S'%'1;
end