Add files via upload

Crucial/Bayesian_Clement.m

@@ -0,0 +1,12 @@
+function DupdateK=Bayesian_Clement(hypps,N,f,clfx,ytrain,oldfolder,ytrue,alpha,combo1)
+% ytrain=y_train;
+Sim=zeros(2,N);
+for i=1:N
+    aa=(hypps(:,i))';
+Sim(:,i)=abs((Forwarding(aa,f,clfx,ytrain,oldfolder,combo1))');
+end
+
+% ytrue=y_train(5,:);
+
+[DupdateK] = ESMDA (hypps,ytrue', N, Sim,alpha);
+end

Crucial/Bayesian_Clement_2.m

@@ -0,0 +1,16 @@
+function DupdateK=Bayesian_Clement_2(hypps,N,f,clfx,ytrain,...
+    oldfolder,ytrue,alpha,combo1,suni)
+% ytrain=y_train;
+%Sim=zeros(2,N);
+parfor i=1:N
+    aa=(hypps(:,i));
+	aa=reshape(aa,[],suni)
+	spit=abs((Forwarding(aa,f,clfx,ytrain,oldfolder,combo1)));
+	spit=reshape(spit,[],1);
+Sim(:,i)=spit;
+end
+
+% ytrue=y_train(5,:);
+
+[DupdateK] = ESMDA (hypps,reshape(ytrue,[],1), N, Sim,alpha);
+end

Crucial/ESMDA.m

@@ -0,0 +1,55 @@
+function [DupdateK] = ESMDA (sgsim,f, N, Sim1,alpha)
+sd=1;
+rng(sd)
+
+%-----------------------------------------------------------------------------    
+
+parfor i=1:size(f,1)
+       aa=f(i,:);
+    aa1=num2str(aa);
+    usee=str2num(aa1(1));    
+   stddWOPR1 =1 %0.01*f(i,:);
+   stdall(i,:)=stddWOPR1;
+   end
+
+nobs = length(f);
+noise = randn(max(10000,nobs),1);
+
+Error1=stdall;
+sig=Error1;
+parfor i = 1 : length(f)
+           f(i) = f(i) + sig(i)*noise(end-nobs+i);
+end
+R = sig.^2;
+  Dj = repmat(f, 1, N);
+           parfor i = 1:size(Dj,1)
+             rndm(i,:) = randn(1,N); 
+             rndm(i,:) = rndm(i,:) - mean(rndm(i,:)); 
+             rndm(i,:) = rndm(i,:) / std(rndm(i,:));
+             Dj(i,:) = Dj(i,:) + sqrt(alpha)*sqrt(R(i)) * rndm(i,:);
+           end
+
+
+Cd2 =diag(R);
+overall=sgsim;
+
+Y=overall; %State variable,it is important we include simulated measurements in the ensemble state variable
+M = mean(Sim1,2);
+% Mean of the ensemble state
+M2=mean(overall,2);
+%M=M'
+% Get the ensemble states pertubations
+parfor j=1:N
+    S(:,j)=Sim1(:,j)-M;
+end
+parfor j=1:N
+    yprime(:,j)=overall(:,j)-M2;
+end
+Cyd=(yprime*S')./((N-1));
+Cdd=(S*S')./((N-1));
+
+Ynew=Y+(Cyd*pinv2((Cdd+(alpha.*Cd2))))*(Dj-Sim1);
+
+DupdateK=Ynew;
+
+end

Crucial/Forwarding.m

@@ -0,0 +1,105 @@
+function [Hardmean,Softmean]=Forwarding(parameters,fv,...
+    clfx,y_train,oldfolder,combo1)
+
+X_test=(clfx.transform(parameters));
+
+cd(oldfolder);
+
+%%
+for ii=1:size(y_train,2)
+% fprintf('Predicting measurement %d | %d .\n', ii,2);   
+folderk = fv; 
+switch combo1
+    case 1
+
+cd(folderk)    
+
+Classifiers=load('Classifiers'); 
+Classifiers=Classifiers.Classifiers;
+
+Experts=load('Experts.out');
+
+
+clfysses=load('clfysses'); 
+clfysses=clfysses.clfysses;
+
+% clfy = MinMaxScalery();
+% (clfy.fit(y_train(:,ii)));
+
+
+Regressors=load('Regressors');
+Regressors=Regressors.Regressors;  
+
+
+Xtrainbig=load('Xtrainbig');
+Xtrainbig=Xtrainbig.Xtrainbig;
+
+ytrainbig=load('ytrainbig');
+ytrainbig=ytrainbig.ytrainbig;
+
+[Hardmean(:,ii)]=prediction_1(Regressors{ii,1},...
+    pred_class(X_test, Classifiers{ii,1})...
+,X_test,Xtrainbig{ii,1},ytrainbig{ii,1},Experts(ii,1),clfysses{ii,1});
+
+[Softmean(:,ii),~]=Unseen_soft_1(Regressors{ii,1},...
+    Classifiers{ii,1},X_test,Xtrainbig{ii,1},ytrainbig{ii,1}...
+,Experts(ii,1),clfysses{ii,1});
+
+    case 2
+
+cd(folderk)       
+
+% disp('Expert=DNN, Gate=DNN')
+Classifiers=load('Classifiers'); 
+Classifiers=Classifiers.Classifiers;
+
+Experts=load('Experts.out');
+
+Classallsbig=load('Classallsbig');
+Classallsbig=Classallsbig.Classallsbig;
+
+clfysses=load('clfysses'); 
+clfysses=clfysses.clfysses;
+
+Regressors=load('Regressors');
+Regressors=Regressors.Regressors;  
+
+% cd(oldfolder)
+[Hardmean(:,ii)]=prediction_2(Regressors{ii,1},pred_class(X_test...
+    , Classifiers{ii,1})...
+,X_test,Classallsbig{ii,1},Experts(ii,1),clfysses{ii,1});
+
+[Softmean(:,ii)]=Unseen_soft_2(Regressors{ii,1},...
+    Classifiers{ii,1},X_test,Classallsbig{ii,1},Experts(ii,1),...
+clfysses{ii,1});
+    case 3
+
+cd(folderk)
+Classifiers=load('Classifiers'); 
+Classifiers=Classifiers.Classifiers;
+
+Experts=load('Experts.out');
+
+Classallsbig=load('Classallsbig');
+Classallsbig=Classallsbig.Classallsbig;
+
+
+clfysses=load('clfysses'); 
+clfysses=clfysses.clfysses;
+
+Regressors=load('Regressors');
+Regressors=Regressors.Regressors;  
+
+% cd(oldfolder)
+[Hardmean(:,ii)]=prediction_3(Regressors{ii,1}...
+    ,str2double(predict(Classifiers{ii,1},X_test))...
+,X_test,Classallsbig{ii,1},Experts(ii,1),clfysses{ii,1});
+
+[Softmean(:,ii),~]=Unseen_soft_3(Regressors{ii,1}...
+    ,Classifiers{ii,1},X_test,Experts(ii,1),clfysses{ii,1});    
+%     
+end
+ cd(oldfolder)
+end
+
+end

Crucial/Forwarding_prime.m

@@ -0,0 +1,108 @@
+function [Hardmean,Softmean]=Forwarding_prime(parameters,...
+    fv,clfx,y_train,oldfolder,combo1)
+X_test=(parameters);
+cd(oldfolder);
+
+%%
+% Answerhard=cell(2,1);
+% Answersoft=cell(2,1);
+for ii=1:2
+% fprintf('Predicting measurement %d | %d .\n', ii,2);   
+folderk = fv; 
+
+switch combo1
+    case 1
+
+cd(folderk)    
+
+
+Classifiers=load('Classifiers'); 
+Classifiers=Classifiers.Classifiers;
+
+Experts=load('Experts.out');
+
+
+clfysses=load('clfysses'); 
+clfysses=clfysses.clfysses;
+
+% clfy = MinMaxScalery();
+% (clfy.fit(y_train(:,ii)));
+
+
+Regressors=load('Regressors');
+Regressors=Regressors.Regressors;  
+
+
+Xtrainbig=load('Xtrainbig');
+Xtrainbig=Xtrainbig.Xtrainbig;
+
+ytrainbig=load('ytrainbig');
+ytrainbig=ytrainbig.ytrainbig;
+
+[Hardmean(:,ii),Hardvariance(:,ii)]=prediction_1(Regressors{ii,1},...
+    pred_class(X_test, Classifiers{ii,1})...
+,X_test,Xtrainbig{ii,1},ytrainbig{ii,1},Experts,clfysses{ii,1});
+[Softmean(:,ii),Softvariance(:,ii)]=Unseen_soft_1(Regressors{ii,1},...
+    Classifiers{ii,1},X_test,Xtrainbig{ii,1},ytrainbig{ii,1}...
+,Experts,clfysses{ii,1});
+
+    case 2
+
+cd(folderk)       
+
+% disp('Expert=DNN, Gate=DNN')
+Classifiers=load('Classifiers'); 
+Classifiers=Classifiers.Classifiers;
+
+Experts=load('Experts.out');
+
+Classallsbig=load('Classallsbig');
+Classallsbig=Classallsbig.Classallsbig;
+
+clfysses=load('clfysses'); 
+clfysses=clfysses.clfysses;
+
+Regressors=load('Regressors');
+Regressors=Regressors.Regressors;  
+
+% cd(oldfolder)
+[Hardmean(:,ii)]=prediction_2(Regressors{ii,1},pred_class(X_test...
+    , Classifiers{ii,1})...
+,X_test,Classallsbig{ii,1},Experts,clfysses{ii,1});
+
+[Softmean(:,ii)]=Unseen_soft_2(Regressors{ii,1},...
+    Classifiers{ii,1},X_test,Classallsbig{ii,1},...
+Experts,clfysses{ii,1});
+    case 3
+
+cd(folderk)
+Classifiers=load('Classifiers'); 
+Classifiers=Classifiers.Classifiers;
+
+Experts=load('Experts.out');
+
+Classallsbig=load('Classallsbig');
+Classallsbig=Classallsbig.Classallsbig;
+
+
+clfysses=load('clfysses'); 
+clfysses=clfysses.clfysses;
+
+Regressors=load('Regressors');
+Regressors=Regressors.Regressors;  
+
+% cd(oldfolder)
+[Hardmean(:,ii),Hardvariance(:,ii)]=prediction_3(Regressors{ii,1}...
+    ,str2double(predict(Classifiers{ii,1},X_test))...
+,X_test,Classallsbig{ii,1},Experts,clfysses{ii,1});
+
+[Softmean(:,ii),Softvariance(:,ii)]=Unseen_soft_3(Regressors{ii,1}...
+    ,Classifiers{ii,1},X_test,Experts,clfysses{ii,1});    
+%     
+end
+ cd(oldfolder)
+end
+
+
+
+end

Crucial/Get_cov.m

@@ -0,0 +1,4 @@
+function cov_big= Get_cov(X_test2,hyp_updated)
+for ik=1:size(X_test2,2)
+cov_big(:,ik)=cov(hyp_updated(:,ik));
+end

Crucial/Get_ensemble.m

@@ -0,0 +1,17 @@
+function ensemble_ini=Get_ensemble(N,X_test2,meanss2,meanss)
+%  for j=1:N
+%     for jj=1:size(X_test2,2)
+%     hyp_inipure(:,jj) = normrnd( meanss2(:,jj),0.1*meanss2(:,jj),1,1) ; 
+%     end 
+%     hyp_inipure=abs(hyp_inipure);
+%     hyp_inipure(:,6)=round(hyp_inipure(:,6));
+%     hyp_inipure(:,8)=round(hyp_inipure(:,8));
+%     hyp_ini=(clfx.transform(hyp_inipure));   
+%     ensemble_ini(:,j)=hyp_ini';
+%  end
+
+ for jj=1:size(X_test2,2)
+  hyp_inipuree(1:N,jj) = unifrnd(meanss2(:,jj),meanss(:,jj),10,1);     
+ end
+ensemble_ini=hyp_inipuree';
+end

Crucial/Get_ensemble_2.m

@@ -0,0 +1,13 @@
+function ensemble_ini=Get_ensemble_2(N,X_test2,meanss2,meanss,Nop)
+
+p=2;
+szss=Nop;
+for k=1:N
+ for jj=1:size(X_test2,2)
+ aj=meanss2(:,jj)+ (meanss(:,jj)- meanss2(:,jj))*sum(rand(szss,p),2)/p;
+ % hyp_inipuree(:,jj) = unifrnd(meanss2(:,jj),meanss(:,jj),szss,1);    
+   hyp_inipuree(:,jj) = reshape(aj,[],1);    
+end
+
+ensemble_ini(:,k)=reshape(hyp_inipuree,[],1);
+end