Copy matrix data (with respect to the given pattern) to xall,
only data from the lower triangle and on the pattern will be used (the rest
is ignored). Quite slow but used only during initialization (e.g., Yinit).
Be careful that here Y can be just part of full length obj.Y,
imagine obj.Yinit{2}=...; then Y{1}=[], Y{2}=... and there is no Y{k>=3} !
Ydefault specifies what should be the default value if Y{k} is not present
on input or it is an empty matrix, possibilities:
-/+Inf/0 ... vector of all -/+Inf/0
anything else (or not present) ... the current point
Y{k} is empty or not present at all --> default value
is of the same size as internal Y --> only lower triangle elements
matching the pattern the internal Y will be considered
is a scalar --> this value will be distributed to all Y in the pattern
(eg. acts as if you put Y{k}=scalar*ones(size(Ymap{k}))0001 % Copy matrix data (with respect to the given pattern) to xall, 0002 % only data from the lower triangle and on the pattern will be used (the rest 0003 % is ignored). Quite slow but used only during initialization (e.g., Yinit). 0004 % 0005 % Be careful that here Y can be just part of full length obj.Y, 0006 % imagine obj.Yinit{2}=...; then Y{1}=[], Y{2}=... and there is no Y{k>=3} ! 0007 % Ydefault specifies what should be the default value if Y{k} is not present 0008 % on input or it is an empty matrix, possibilities: 0009 % -/+Inf/0 ... vector of all -/+Inf/0 0010 % anything else (or not present) ... the current point 0011 % Y{k} is empty or not present at all --> default value 0012 % is of the same size as internal Y --> only lower triangle elements 0013 % matching the pattern the internal Y will be considered 0014 % is a scalar --> this value will be distributed to all Y in the pattern 0015 % (eg. acts as if you put Y{k}=scalar*ones(size(Ymap{k})) 0016 function vec = Y2vec(obj, Y, Ydefault) 0017 0018 % use a point to fill in the gabs if there are in Y 0019 % if Y is a full size array, it won't influence it anyway 0020 % important only for Y being partially assigned 0021 if (nargin<=2) 0022 % current point 0023 vec=obj.xall(obj.Nx+1:end); 0024 else 0025 if (Ydefault==-Inf) 0026 vec=-Inf(obj.NYnnz,1); 0027 elseif (Ydefault==Inf) 0028 vec=Inf(obj.NYnnz,1); 0029 elseif (Ydefault==0) 0030 vec=zeros(obj.NYnnz,1); 0031 else 0032 vec=obj.xall(obj.Nx+1:end); 0033 end 0034 end 0035 0036 if (length(Y)>obj.NY) 0037 error('Incompatible Y on input, cannot transfer data'); 0038 end 0039 0040 for k=1:length(Y) 0041 if (~isempty(Y{k})) 0042 mapper=obj.vec2Ymap{k}; 0043 0044 [dim, dim2] = size(Y{k}); 0045 if (dim==1 && dim2==1) 0046 % scalar --> apply this bound on all elements 0047 vec(mapper.xmap)=Y{k}; 0048 0049 elseif (dim==dim2 && dim==mapper.dim) 0050 % matching dimension, get the appropriate elements 0051 0052 if (mapper.dense) 0053 % mapper.xmap includes double assignments (from both triangles) 0054 % to be sure which half of the user's data gets used, symmetrize 0055 % and use lower only 0056 0057 Ytmp=tril(Y{k}) + tril(Y{k},-1)'; 0058 vec(mapper.xmap)=reshape(full(Ytmp),dim*dim,1); 0059 else 0060 len=length(mapper.xmap); 0061 % have to do it element by element because MATRIX(irow,icol) will 0062 % be a submatrix not an array of arguments. Find is also not suitable 0063 % because it won't register zeros (which are on the pattern of variables) 0064 for idx=1:len 0065 ridx=mapper.irow(idx); 0066 cidx=mapper.icol(idx); 0067 if (ridx>=cidx) 0068 % lower triangle only 0069 vec(mapper.xmap(idx))=Y{k}(ridx,cidx); 0070 end 0071 end 0072 end 0073 else 0074 % not matching dimension and not a scalar (dim~=dim2 || dim~=mapper.dim) 0075 error('Y{k} has different size.'); 0076 end 0077 end 0078 end 0079 0080 0081