Group-FOS/script_simulation_logFOS.m at master · LedererLab/Group-FOS · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
% addpath(genpath('../glmnet'))
%clear all;
%% Simulation of data from logistic regression models
nObs=200; %number of data examples
nVarsVec=200;% number of variables
nVarsVecSize = length(nVarsVec);
nActive=8;  %number of activated variables
method='logFOS';
corVec=0.5;

corVecSize = length(corVec);
signalNoiseRatio=5;
signallevel=2;
cStats =6; % c_log in the paper
cComp = 1; % z in the paper


fracMax = 1e-4; % fraction of largest tuning parameter
% Careful ! standardize = false if data already standardized
options_glmnet.standardize = false;
options_glmnet.intercept = false;
% Number of folds for cross-validation
Kfold = 10;
nRun = 10;
% Hamming distance means and standard deviations

hamming_dist_FOS_ista_mean_std = zeros(nVarsVecSize*corVecSize,2);
hamming_dis_LI_std=zeros(nVarsVecSize*corVecSize,2);
hamming_dis_lassospams_std=zeros(nVarsVecSize*corVecSize,2);
% Run time means and standard deviations

time_FOS_ista_mean_std = zeros(nVarsVecSize*corVecSize,2);
time_LI_std=zeros(nVarsVecSize*corVecSize,2);
time_lassospams_std=zeros(nVarsVecSize*corVecSize,2);

estim_FOS_ista_mean_std = zeros(nVarsVecSize*corVecSize,2);
estim_log_lasso_mean_std=zeros(nVarsVecSize*corVecSize,2);
estim_LI_mean_std=zeros(nVarsVecSize*corVecSize,2);

for iVars=1:nVarsVecSize
    for iCor=1:corVecSize


        hamming_dist_FOS_ista_tmp = zeros(1, nRun);
        hamming_dist_LI_tmp= zeros(1, nRun);
        hamming_dist_lassospams_tmp= zeros(1, nRun);


        time_FOS_ista_tmp = zeros(1, nRun);
        time_LI_tmp=zeros(1,nRun);
        time_lassospams_tmp=zeros(1,nRun);

        estim_Log_FOS = zeros(1, nRun);
        estim_LI = zeros(1, nRun);
        estim_Log_lasso = zeros(1, nRun);
        for iRun=1:nRun
            rng((iRun+3));
            [X_unstd, y_unstd, betaTrue_unstd] = simulate_logistic_regression(nObs,nVarsVec(iVars),nActive,corVec(iCor),signallevel);
            suppTrue = (betaTrue_unstd ~= 0);
            X=X_unstd;
            y=y_unstd;
            lambda_max = 10*(log(nVarsVec)/nObs);
             %lambda_max=0.001;
             lambdaVec_FOS = sort(linspace(fracMax*lambda_max, lambda_max, 500), 'descend');
             tic
            [betaFOS,lambdaFOS,suppFOSind] = FOS_log_reg(X,y,lambdaVec_FOS,cStats,cComp);
	        time_FOS_ista_tmp(iRun) = toc;
            suppFOS = zeros(nVarsVec(iVars), 1);
	        suppFOS(suppFOSind) = 1;
            hamming_dist_FOS_ista_tmp(iRun) = sum(suppFOS ~= suppTrue);
            estim_Log_FOS(iRun)=norm(betaFOS-betaTrue_unstd,Inf);


%%%%%%%%%%%%loglassoSpams
          varError = 1;
          tic;
%
%           [beta_CV] = cv_loglasso_fista_v2(X,y,lambdaVec_FOS,Kfold);
           beta_CV=zeros(nVarsVec(iVars),1);
          time_lassospams_tmp(iRun) = toc;
          supplassospams = (beta_CV~=0);
          hamming_dist_lassospams_tmp(iRun) = sum(supplassospams ~= suppTrue);
           estim_Log_lasso(iRun)=norm(beta_CV-betaTrue_unstd,Inf);
%%%%%%%%%%%%%%%%%%% LI2019
%             tic
%             [betaLI,lambdaLI,suppLIindLI] = Li_log_reg(X,y,lambdaVec_FOS,cStats,cComp);
% 	        time_LI_tmp(iRun) = toc;
%             suppLI = zeros(nVarsVec(iVars), 1);
% 	        suppLI(suppLIindLI) = 1;
%             hamming_dist_LI_tmp(iRun) = sum(suppLI ~= suppTrue);
%             estim_LI(iRun)=norm(betaLI-betaTrue_unstd,Inf);
             end

        % Hamming distance means

         hamming_dist_FOS_ista_mean_std(corVecSize*(iVars-1)+iCor, 1) = mean(hamming_dist_FOS_ista_tmp);
%          hamming_dis_LI_std(corVecSize*(iVars-1)+iCor, 1) = mean( hamming_dist_LI_tmp);
         hamming_dis_lassospams_std(corVecSize*(iVars-1)+iCor, 1) = mean( hamming_dist_lassospams_tmp);


        % Hamming distance standard deviations
         hamming_dist_FOS_ista_mean_std(corVecSize*(iVars-1)+iCor, 2) = std(hamming_dist_FOS_ista_tmp);
%          hamming_dis_LI_std(corVecSize*(iVars-1)+iCor, 2) = std( hamming_dist_LI_tmp);
         hamming_dis_lassospams_std(corVecSize*(iVars-1)+iCor, 2) = std(hamming_dist_lassospams_tmp);

        % Mean run times
         time_FOS_ista_mean_std(corVecSize*(iVars-1)+iCor, 1) = mean(time_FOS_ista_tmp);
%          time_LI_std(corVecSize*(iVars-1)+iCor, 1) = mean(time_LI_tmp);
         time_lassospams_std(corVecSize*(iVars-1)+iCor, 1) = mean(time_lassospams_tmp);

        % Standard deviation run times
         time_FOS_ista_mean_std(corVecSize*(iVars-1)+iCor, 2) = std(time_FOS_ista_tmp);
%          time_LI_std(corVecSize*(iVars-1)+iCor, 2) = std(time_LI_tmp);
         time_lassospams_std(corVecSize*(iVars-1)+iCor, 2) = std(time_lassospams_tmp);

         estim_FOS_ista_mean_std (corVecSize*(iVars-1)+iCor, 1) = mean(estim_Log_FOS);
         estim_log_lasso_mean_std (corVecSize*(iVars-1)+iCor, 1) = mean(estim_Log_lasso);
         estim_FOS_ista_mean_std (corVecSize*(iVars-1)+iCor, 2) = std(estim_Log_FOS);
         estim_log_lasso_mean_std (corVecSize*(iVars-1)+iCor, 2) = std(estim_Log_lasso);
%          estim_LI_mean_std (corVecSize*(iVars-1)+iCor, 2) = std(estim_LI);
%          estim_LI_mean_std (corVecSize*(iVars-1)+iCor, 1) = mean(estim_LI);


    end
end
sdate_str = date;

fprintf('\n log-Fos_HamDis=%4.2f, Var_HamDis=%4.2f', hamming_dist_FOS_ista_mean_std(1,1), hamming_dist_FOS_ista_mean_std(1,2))
fprintf('\n log-Fos_EstErr=%4.2f, Var_EstErr=%4.2f',  estim_FOS_ista_mean_std (1,1), estim_FOS_ista_mean_std (1,2))
fprintf('\n log-Fos_RunTime=%4.2f, Var_RunTime=%4.2f ',  time_FOS_ista_mean_std(1,1),time_FOS_ista_mean_std(1,2))