Add files via upload

Author: KKahirman

ComparisonWeakSINDy/Main_NoiseCompare01.m

@@ -0,0 +1,106 @@
+%% This file will swipe the noise level
+clc;clear all;close all;
+%% First load the noise and simulation data
+load('GeneratedNoise.mat')
+addpath("Result")
+addpath("Functions")
+%% Define the parameters for the WeakSINDy_AG
+% Library order
+polyorder=2;
+usesine=0;
+
+% Optimization parameter
+gamma=0;
+tau = 1;
+
+% Number of test function
+K = 200;
+
+% Test function order
+poly = 14; 
+
+% Test function support size
+s = 31;
+r_whm = 8;
+
+% Parameters for the WeakSINDy
+wsindy_params = {s, K, poly, tau};
+
+% Define the thresholding parameters
+lam=linspace(0.01,0.95,30);
+lamNum=length(lam);
+
+% Define a matrix to store the values
+Theta_test=poolData(x_test,stateVar,polyorder,usesine);
+Xi0=zeros(lamNum,size(Theta_test,2),stateVar);
+TestingScore=zeros(lamNum,1);
+SuccessOrNot_WS=zeros(N_run,NoiseNum);
+Xi_WS=zeros(lamNum,size(Theta_test,2),stateVar);
+
+% Store the parameter error
+Eparm_WS=zeros(N_run,NoiseNum);
+
+Xi_base=[-10,28,0;
+         10,-1,0;
+         0,0,-8/3;
+         0,0,0;
+         0,0,1;
+         0,-1,0;
+         0,0,0;
+         0,0,0;
+         0,0,0];
+%% Next run a for loop to test each noise level
+pin=0;
+
+for i=1:N_run
+    fprintf(strcat("This is the run:",string(i),"\n"))
+    for j=1:NoiseNum
+        fprintf(strcat("\t","Using noise level as: ",string(NoisePercentageToSwipe(j)),"\n"))
+        pin=pin+1;
+        % Get the observation data
+        xes=squeeze(xes_List(pin,:,:));
+        Theta=poolData(xes,stateVar,polyorder,usesine);
+        
+        % Swipe the thresholding parameter
+        parfor k=1:lamNum
+            fprintf(strcat("\t\t","Swiping lambda, using lambda as: ",string(lam(k)),"\n"))
+            
+            % Get the result for current lam
+            Xi0(k,:,:)=WeakSINDy_AG(t,xes,Theta,stateVar,lam(k),gamma,r_whm,wsindy_params);
+            
+            % Calculate the error on the test data
+            TestingScore(k,1)=norm(dx_test-Theta_test*squeeze(Xi0(k,:,:)))/norm(dx_test);
+            
+        end
+        
+        % Test whether the structure of the model is correct
+        for k=1:lamNum
+            if norm((Xi_base~=0)-(squeeze(Xi0(k,:,:))~=0))==0
+                SuccessOrNot_WS(i,j)=1;
+                Xi_dummy=squeeze(Xi0(k,:,:));
+                Xi_WS(pin,:,:)=Xi_dummy;
+            end
+        end
+        
+        % If no model has the correct structure, choose the model that has
+        % lowest testing error as final model
+        if SuccessOrNot_WS(i,j)==0
+           % Now select the best Xi
+            [minVal,index]=min(TestingScore);
+        
+            % Store the current result
+            Xi_dummy=squeeze(Xi0(index,:,:));
+            Xi_WS(pin,:,:)=Xi_dummy;
+        end
+        
+        % Store the parameter error
+        Eparm_WS(i,j)=norm(Xi_base-Xi_dummy)/norm(Xi_base);
+        
+    end
+end
+
+%% Save the results
+Ep_WS_01=Eparm_WS;
+SuccessOrNot_WS_01=SuccessOrNot_WS;
+
+%save ("Result\W_SINDy_dt_01.mat","Ep_WS_01","SuccessOrNot_WS_01")