MVR Composer Code

function CondNumErr();
%%used alexey's code
warning off;%
global PLTOPTSFIGNUM;
PLTOPTSFIGNUM = 1;
THISFOLDER = fileparts(mfilename('fullpath'));
CODEFOLDER = fullfile(THISFOLDER,'code');
REPORTFOLDER  = fullfile(THISFOLDER,'report');
LASSOFOLDER = fullfile(THISFOLDER,'lasso');
SWFOLDER = fullfile(THISFOLDER,'fsbox');
DATAFOLDER = fullfile(THISFOLDER,'data');
if nargin < 1, Genname = 'gen.mat'; end
GenFullFname = fullfile(DATAFOLDER, Genname);
load(GenFullFname);
if ~exist(REPORTFOLDER,'dir'), mkdir(REPORTFOLDER); end
addpath(CODEFOLDER);
addpath(SWFOLDER);
addpath(LASSOFOLDER);
%if nargin < 1, PrjFname = 'med.prj.txt'; end

if nargin < 1, PrjFname = 'rom.prj.txt'; end
PrjFullFname = fullfile(DATAFOLDER, PrjFname);
% read the project file 
evalstr = textread(PrjFullFname,'%s','commentstyle','matlab','delimiter','%');
for i=1:length(evalstr)
    try 
        eval(evalstr{i});   
        fprintf(1,'%s\n',evalstr{i});
    catch
        fprintf(1,'\nError in the project file %s, line = %d, content = %s \n', PrjFname, i, evalstr{i});
    end
end 


AverNum = 10;
CondNum = [];

n = 150; % number of objects
m =20;  % number of varibles
errTest = zeros(length(alg),6);
errLearn = zeros(length(alg),6);

Cond = zeros(length(alg),11);
%legend

for j = 1:length(alg)
    CondNum = [];
    %errTest = [];
    %errLearn = [];
    p=0;
    for e=0:0.1:1 % e is conditionality measure, not conditionality
    p = p+1;
% generate model data    
        X1  = e*ones(n,m-5) +(1-e)*[diag(ones(m-5,1));zeros(n-m+5,m-5)];
        
        %X(:,m/2+1:end) = X(:,m/2+1:end)+X(:,1:m/2);
        X = [X1 eye(n,5)];%; zeros(20,1)]);zeros(n-m,m)];
        %X1  = ones(n,m) -[diag(ones(m,1));zeros(n-m,m)];
        %X = e*ones(n,m)+ (1-e)* [diag(ones(m,1));zeros(n-m,m)];      
        %beta = (rand(m,1)-0.5); % linear regression parameters
        beta = [zeros(m-10,1); ones(10,1)];%ones(5,1);
        y = X*beta;%MN; % get responses
        %y = y + (rand(n,1)-0.5)/10;
        X = X + randMN(:,1:m);%(rand(n,m) - 0.5)/10; % add noise 

        [U,L,V]=svd(X);% find condition number for current XA
        r = min(size(L));
        CondNum(end+1) = L(1,1)/L(r,r);
 
        for i=1:AverNum
            [idxA,idxB]=SplitData(n,TRAIN2TEST);
            yA=y(idxA);
            A=X(idxA,:);
            yB=y(idxB); 
            B=X(idxB,:);
            w = lscov(B,yB);
            mdlopts.sigma = sumsqr(yB - B*w)/size(B,1);% for Cp-estimation
            mdlopts.algoName = alg{j};
            fprintf(1,strcat('\n ', mdlopts.algoName));%Call FOS
            %learn using chosen method
            [b,Qlearn] = eval(strcat(mdlopts.algoName,'( A, yA, mdlopts )'));
            %test
           % figure;
            s1 = sum(b.^2,2).^(0.5);
            plot(s1,b);
            xlabel(texlabel('|beta_i|'));
            ylabel(texlabel('beta_i'));
           % hold off
            [best,Qtest,Qlearn] = test1(A,yA,B,yB,b);
            [Qtest,idx]=sort(Qtest);
            Qlearn = Qlearn(idx);
            b = b(idx,:);
            if ~isempty(b)
                [Pop{i}] = CreatePopulation(B,yB,b,Qtest,Qlearn,mdlopts,n);end
        end
          if ~isempty(Pop)
          [PopMean]=MeanCharact(Pop,mdlopts);end
        %Population2TeX(PopMean,ReportFilename,mdlopts.algoName);
           errLearn(j,p) = PopMean.errTrain;
           errTest(j,p) = PopMean.errTest;
        
       % [U,L,V]=svd(X(:,find(PopMean.wFound)));% find condition number for current XA
       % r = min(size(L));
       % Cond(j,p) = L(1,1)/L(r,r);
    end
    

end
figure;
str = {'.--r' '.-b' '.-g' '.:' '.-m' '.c'};
for j = 1:(length(alg))
     hold on;
    ylabel('Errors');
    xlabel('Condition number');
    %title(namealg(j));
   % plot(CondNum,errLearn,'--red');
   % plot(CondNum,Cond(j,:),str{j});
    
    plot(CondNum,errTest(j,:),str{j});
   % legend('Learn','Control',1)
 %   text(strcat('\leftarrow ',namealg{j}),...
  %   'Ho5rizontalAlignment','left')
    %hold off
end
   legend(str,namealg);
%rmpath(FUNCFOLDER);
rmpath(CODEFOLDER);
return