TODO_list

1.  MakeQuakes: should use quakeopt.file and quakeopt.format. Should introduce this the moment we work with 2 quakefiles.
1.1 Prepare:    data should be loaded using  (requires changes of the *min files and probably messing around with symbopt
             Quakes = MakeQuakes(quakeopt)
             Faults = MakeFaults(faultopt)
             Roads  = MakeRoads (roadopt)
             GPS    = MakeGPS   (gpsopt)

3. Check whether geodmod version is consistent with *min file
cvs status -v README_INSTALL | grep -A1 Existing | tail -1 | awk '{print $1}' >! masterfile/geodmod_version

4.1 do the bw_areaoen thing also after the median filtering in Igram2Motion
    we probably need : makesaropt.igram2motionopt.cormask_area_open    =  500


5. Changing read_from_file so that commands are evaluated
TESTDATA=getenv('GEODMOD_PROJECTDATA')
q1='dirvar'
q2='strcat(TESTDATA,''/dem'')'
eval(sprintf('%s=%s ;',q1,q2))

6. calculation progress:  (the following works well from Loren's Blog)
     % Example Code for Printing Progress to the Command Window
     fprintf(1,'here''s my integer:     ');
     for i=1:999
         fprintf(1,'\b\b\b%3d',i); pause(.2)
     end
     fprintf('\n')
7. Change source description:
inverseopt.source(1).description                    =   {'disloc', 'linear', 'xy'}
inverseopt.source(1).lowerbound                     =  [ 4.0  4.0  1.0  -89.9 25.0  10.0  10.0  0.0  0.0 1.0  ];
inverseopt.source(1).upperbound                     =  [ 4.0  4.0  1.0  -89.9 25.0  10.0  10.0  0.0  0.0 1.0  ];
inverseopt.source(2).description                    =   {'mogi', 'linear', 'xy'}
inverseopt.source(2).lowerbound                     =  [ 4.0  4.0  1.0  1.0  ];
inverseopt.source(2).upperbound                     =  [ 4.0  4.0  1.0  1.0  ];

or
inverseopt.source(1).descriptmin                     =  [ 'disloc', 'fixed', 'xy' ,  4.0  4.0  1.0  -89.9 25.0  10.0  10.0  0.0  0.0 1.0  ];
inverseopt.source(1).descriptmax                     =  [ 'disloc', 'fixed', 'xy' ,  4.0  4.0  1.0  -89.9 25.0  10.0  10.0  0.0  0.0 1.0  ];

8. Check input *min file for consistency.
   I had the case with N_disloc=1 and N_mogi=0 
   but inverseopt.mogi_bounds.ll was given which caused a disaster. It was very difficult to find the problem

9. GridSearch should have an option to produce the model parameters in matrix form for quick model evaluation. The following in
gibbs.m (line 128) does the objective fuction call in matrix form (same code in anneal.m)

			if matrix
				O=feval(FUN,modelmatrix,varargin{:});
			else
				O=zeros(NM,1);
				for e=1:NM
					O(e)=feval(FUN,modelmatrix(:,e),varargin{:});
	                count=count+1;
           		end
			end
modelmatrix is a 3x9 variable for a mogi source with 3 nonlin parameters  

10. Parallelizing anneal. I tried to use a parfor loop over the runs (so that 4 runs are done in parallel) but I get errors because variables
energy and model can't be classified for parfor. Don't know why. the following works:
matlabpool local 4
aa=zeros(8,1);
tic; parfor(i=1:8)   a=rand(5000); aa(i)=max(max(a)); disp(aa(i)); end ; toc
tic; for(i=1:8)   a=rand(5000); aa(i)=max(max(a)); disp(aa(i)); end ; toc

in anneal.m the following made things slower but not faster:
     parfor e=1:NM   
            O(e)=feval(FUN,modelmatrix(:,e),varargin{:});
     end

     parfor x=1:p
gives the error message: the variable count and energy can't be classified