Changeset 3134

Timestamp:

02/25/10 08:25:01 (15 years ago)

Author:

Mathieu Morlighem

Message:

simplified parametercontroloptimization

File:

• issm/trunk/src/m/classes/public/parametercontroloptimization.m (modified) (2 diffs)

issm/trunk/src/m/classes/public/parametercontroloptimization.m

@@ -10 +10 @@
 %
 %   Example:
-%      md=parametercontroloptimization(md,'nsteps',30)
+%      md=parametercontroloptimization(md,'nsteps',6)
 
 %process options
@@ -21 +21 @@
 %variable
 cmax_max=1000;
+cm_maxs=linspace(cmax_max/10,cmax_max,3);
 optscal_max=1000;
+optscals=linspace(optscal_max/10,optscal_max,3);
+fits=[0 2];
 %Kept constant
+md2.verbose=0;
 md2.cm_min=1;
-md2.nsteps=5;
+md2.eps_cm=NaN;
 md2.cm_noisedmp=0;
 md2.cm_maxdmp_slope=0;
 md2.cm_mindmp_slope=0;
-md2.eps_cm=NaN;
+md2.nsteps=getfieldvalue(options,'nsteps',5);
+md2.control_type=getfieldvalue(options,'md2.control_type','drag');
 md2.maxiter=10*ones(md2.nsteps,1);
 md2.cm_jump=0.99*ones(md2.nsteps,1);
-md2.verbose=0;
 md2.control_analysis=1;
-md2.control_type=getfieldvalue(options,'md2.control_type','drag');
 
-%Get nsteps
-nsteps=getfieldvalue(options,'nsteps',50);
-if mod(nsteps,5)~=0,
-        error('''nsteps'' option should be a multiple of 5')
-end
+%loop over the set of parameters
+best=0;
+count=1;
+for fit=fits,
+        md2.fit=fit*ones(md2.nsteps,1);
+        for cm_max=cm_maxs;
+                md2.cm_max=cm_max;
+                for optscal=optscals;
+                        md2.optscal=optscal*ones(md2.nsteps,1);
 
-%loop over the set of 5 steps
-fit_final=[];
-cmmax_final=[];
-optscal_final=[];
-for i=1:nsteps/5,
-        disp(['Step ' num2str(i) '/' num2str(floor(nsteps/5))]);
-        best=0;
-        for fit=[0 2],
-                md2.fit=fit*ones(md2.nsteps,1);
-                for cm_max=linspace(cmax_max/10,cmax_max,3),
-                        md2.cm_max=cm_max;
-                        for optscal=linspace(optscal_max/10,optscal_max,3),
-                                md2.optscal=optscal*ones(md2.nsteps,1);
+                        %current run
+                        disp(sprintf('\n   Step %i/%i, fit=%i, cm_max=%g, optscal=%g\n',count,length(cm_maxs)*length(optscals)*length(fits),fit,cm_max,optscal))
+                        md2=solve(md2,'analysis_type','diagnostic');
+                        count=count+1;
 
-                                %current run
-                                md2=solve(md2,'analysis_type','diagnostic');
-
-                                %Check misfit difference
-                                rel=(md2.results.diagnostic.J(1)-md2.results.diagnostic.J(end))/md2.results.diagnostic.J(1);
-                                disp(['   ΔJ/J=' num2str(rel*100) '% for fit=' num2str(fit) ',cm_max=' num2str(cm_max) ', and optscal=' num2str(optscal) ]);
-                                if (rel>best),
-                                        disp(['   --> best for this step']);
-                                        fit_final(:,i)    =md2.fit;
-                                        cmmax_final(:,i)  =md2.cm_max;
-                                        optscal_final(:,i)=md2.optscal;
-                                        drag=md2.results.diagnostic.parameter;
-                                        best=rel;
-                                end
+                        %Check misfit difference
+                        rel=(md2.results.diagnostic.J(1)-md2.results.diagnostic.J(end))/md2.results.diagnostic.J(1);
+                        disp(['   ΔJ/J=' num2str(rel*100) '% for fit=' num2str(fit) ',cm_max=' num2str(cm_max) ', and optscal=' num2str(optscal) ]);
+                        if (rel>best),
+                                disp(sprintf('   --> current best'))
+                                fit_final    =md2.fit;
+                                cmmax_final  =md2.cm_max;
+                                optscal_final=md2.optscal;
+                                best=rel;
                         end
                 end
         end
-        %Plug best drag;
-        md2.drag=drag;
 end
 
 %load final parameters onto initial model
-md.fit    =fit_final(:);
-md.cm_max =cmmax_final(:);
-md.optscal=optscal_final(:);
+md.fit    =fit_final;
+md.cm_max =cmmax_final;
+md.optscal=optscal_final;