In Matlab 2012b, there is a changem function that allows you to substitute elements of a matrix with other values specified by a set of keys:
Substitute values in data array
Is there an elegant/vectorized way to do the same if I don't have the Mapping toolbox?
Yes, use ismember:
A = magic(3);
oldCode = [ 8 9];
newCode = [12 13];
[a,b] = ismember(A,oldCode);
A(a) = newCode(b(a));
I don't know changem, and I suspect the above will not fully cover its functionality (why else would TMW have introduced changem?), but well, it does what you asked :)
Vectorized implementation of CHANGEM with bsxfun, max
Sometime back, I was made to write a customized vectorized version of changem implemented with bsxfun and max as part of a much bigger problem. The referenced solution could be found here. Then, following few links I saw this post and thought it could be posted here as a solution for an easy find among future readers as also because this problem exclusively asks for an efficient and vectorized version of changem. So, here's the function code -
%// CHANGEM_VECTORIZED Vectorized version of CHANGEM with MAX, BSXFUN
function B = changem_vectorized(A,newval,oldval)
B = A;
[valid,id] = max(bsxfun(@eq,A(:),oldval(:).'),[],2); %//'
B(valid) = newval(id(valid));
return;
The syntax used in the custom version follows the same syntax as in changem.m -
function B = changem(A, newval, oldval)
%CHANGEM Substitute values in data array ...
Unfortunately, I think you need a FOR loop. But it's pretty straightforward:
function xNew = myChangeM(x,oldCode,newCode)
% xNew = myChangeM(x,oldCode,newCode)
%
% x is a matrix of vaues
% oldCode and newCode specify the values to replace and with what
% e.g.,
% x = round(randn(10));
% oldCode = [-1 -2];
% newCode = [nan, 10]; %replace -1 with nan, -2 by 10
% xNew = myChangeM(x,oldCode,newCode)
xNew = x;
for repInd = 1:numel(oldCode)
xNew(x == oldCode(repInd)) = newCode(repInd);
end
