calc_meas_icov

 meas_icov = calc_meas_icov( inv_model )
 CALC_MEAS_ICOV: calculate inverse covariance of measurements
   The meas_icov is a matrix n_meas x n_meas of the
     inverse of measurement covariances. Normally measurements
     are assumed to be independant, so the meas_icov is 
     a diagonal matrix of 1/var(meas)

 calc_meas_icov can be called as
    meas_icov= calc_meas_icov( inv_model )

 meas_icov   is the calculated data prior
 inv_model    is an inv_model structure

 if:
    inv_model.meas_icov    is a function
          -> call it to calculate meas_icov
    inv_model.meas_icov    is a matrix
          -> return it as meas_icov
    inv_model.meas_icov    does not exist
          -> use I, or 1./homg (for normalized difference)

0001 function meas_icov = calc_meas_icov( inv_model )
0002 % meas_icov = calc_meas_icov( inv_model )
0003 % CALC_MEAS_ICOV: calculate inverse covariance of measurements
0004 %   The meas_icov is a matrix n_meas x n_meas of the
0005 %     inverse of measurement covariances. Normally measurements
0006 %     are assumed to be independant, so the meas_icov is
0007 %     a diagonal matrix of 1/var(meas)
0008 %
0009 % calc_meas_icov can be called as
0010 %    meas_icov= calc_meas_icov( inv_model )
0011 %
0012 % meas_icov   is the calculated data prior
0013 % inv_model    is an inv_model structure
0014 %
0015 % if:
0016 %    inv_model.meas_icov    is a function
0017 %          -> call it to calculate meas_icov
0018 %    inv_model.meas_icov    is a matrix
0019 %          -> return it as meas_icov
0020 %    inv_model.meas_icov    does not exist
0021 %          -> use I, or 1./homg (for normalized difference)
0022 
0023 % (C) 2005 Andy Adler. License: GPL version 2 or version 3
0024 % $Id: calc_meas_icov.html 2819 2011-09-07 16:43:11Z aadler $
0025 
0026 meas_icov = eidors_obj('get-cache', inv_model, 'meas_icov');
0027 
0028 if ~isempty(meas_icov)
0029    eidors_msg('calc_meas_icov: using cached value', 3);
0030    return
0031 end
0032 
0033 if isfield(inv_model,'meas_icov')
0034    if isnumeric(inv_model.meas_icov);
0035       meas_icov = inv_model.meas_icov;
0036    else
0037       meas_icov= feval( inv_model.meas_icov, inv_model);
0038    end
0039 else
0040    meas_icov= default_meas_icov( inv_model );
0041 end
0042 
0043 eidors_obj('set-cache', inv_model, 'meas_icov', meas_icov);
0044 eidors_msg('calc_meas_icov: setting cached value', 3);
0045  
0046 % Calculate a data prior for an assumption of uniform noise
0047 % on each channel
0048 %
0049 function meas_icov = default_meas_icov( inv_model )
0050 
0051    fwd_model= inv_model.fwd_model;
0052    if isfield(fwd_model,'normalize_measurements')
0053       normalize = fwd_model.normalize_measurements;
0054    else
0055       normalize = 0;
0056    end
0057 
0058    n =  calc_n_meas( fwd_model );
0059 
0060    if ~normalize
0061       meas_icov= speye( n );
0062    else
0063       homg_data=  solve_homg_image( fwd_model );
0064 % sig = k/h -> std = k/h -> 1/std = kh
0065       meas_icov = sparse(1:n, 1:n, ( homg_data.meas ).^2 );
0066    end
0067 
0068 function n_meas = calc_n_meas( fwd_model )
0069 
0070    n_meas = 0;
0071    for i= 1:length(fwd_model.stimulation );
0072        n_meas = n_meas + size(fwd_model.stimulation(i).meas_pattern,1);
0073    end
0074 
0075 % create homogeneous image + simulate data
0076 function homg_data = solve_homg_image( fwd_mdl )
0077     n_elems= size( fwd_mdl.elems , 1);
0078     mat= ones( n_elems, 1);
0079     homg_img= eidors_obj('image', 'homogeneous image', ...
0080                          'elem_data', mat, 'fwd_model', fwd_mdl );
0081     homg_data=fwd_solve( homg_img);