GreedyPlusMILP_Final_1.run

# THIS CODE IS FOR RUNNING MILP ON DATA OBTAINED AFTER APPLYING GREEDY APPROACH
# WE WILL USE THETA BASED EVALUATION (METRIC 1) AND MISSCLASSIFICATION BASED EVALUATION (METRIC 2)
# WE WILL ALSO KEEP TRACK OF COMPUTATION TIME


#RESET THE AMPL ENVIROMENT
reset;


#LOAD THE MODEL
model ReducedOutliersMILPmain.mod; #Use this model if data has no noise;
#model NoisyMILPModel.mod; #Use this model if data has some noise....

#CHANGE THE SOLVER (optional)
option solver gurobi;
option gurobi_options 'timelim 60'; #'bestobjstop 20';



param FP;
param inliers;
param ratio;
param sigma {1..4}; 
param time1;  #Keep track of time;
param time2;
param norm_of_diff;
param norm_theta_hat;
param norm_theta_true;
param metric_1_ratio;



#SOLVE
for {ncase in 1..1} 
{   
    for {pcase in 15..17}
    { 
        for {case in 1..50}
        {     
              
              display ncase;
              display pcase;
              display case;
              reset data;
              option randseed 0;
              data ("./inputMILP/Noise/GreedyPlusMILP/Removed_50_Percent/reducedData_n_" & ncase & "_p_" & pcase & "_" & case & ".dat");
              display p;
      
              
              #Time Calculation           
              let time1 := _total_solve_elapsed_time;
              solve;
              let time2 := _total_solve_elapsed_time;
              display theta, T, X_true, X, outliervec, Z, f;
              display _total_solve_elapsed_time;
              display time2 - time1;
              display greedy_time;
              print (time2 - time1 + greedy_time) >> ("./outputMILP/GreedyPlusMILP/GreedyRemoved_50_Percent/Time/data_n_" & ncase & "_p_" & pcase & ".out");
              
              #Metric_1 based Evaluation
              let norm_of_diff :=0;
              let norm_theta_hat :=0;
              let norm_theta_true :=0;
              
              for {i in 1..r} 
              {
               let norm_of_diff := norm_of_diff + (theta[i] - T[i])*(theta[i] - T[i]);
               let norm_theta_hat := norm_theta_hat + T[i]*T[i];
               let norm_theta_true := norm_theta_true + theta[i]*theta[i];
              }
              
              let metric_1_ratio := norm_of_diff/(norm_theta_hat + norm_theta_true) ;
              display metric_1_ratio;
              print metric_1_ratio >> ("./outputMILP/GreedyPlusMILP/GreedyRemoved_50_Percent/Metric_1/data_n_" & ncase & "_p_" & pcase & ".out");
              
             
              
              # Metric-2 Based Evaluation
              let FP :=0;
              let inliers :=0;
              for {i in 1..D}
              {
                  if(outliervec_original[i] = 0) then {
                       let inliers := inliers + 1;
                  }
              
              }
              
              # These below 4 lines of code are unnecessary, but might be helpful to compute x_hat (estimated x)
              let sigma[1] := 0;
              let sigma[2] := 1e-9;
              let sigma[3] := 1e-3;
              let sigma[4] :=1e-1;
              
            
              
              for { i in 1..d}
              {
                          
                          #It determines false negative; inliers being treated as outliers
                          
                          if(outliervec[i] = 0 && Z[i] = 1) then { 
                                 let FN :=FN+1; 
                          }
                          
                          #It determines false positive; outliers beign treated as inliers..
                          if(outliervec[i] = 1 && Z[i] = 0) then { 
                                 let FP :=FP+1; 
                          }      
              
              }
              let ratio := (FP+FN)/inliers ;
              display inliers, FP, FN, ratio;
              print ratio >> ("./outputMILP/GreedyPlusMILP/GreedyRemoved_50_Percent/Metric_2/reducedData_n_" & ncase & "_p_" & pcase & ".out");
         }
         
      }              
}