Error: Assignment has more non-singleton rhs dimensions than non-singleton subscripts

I have the following matlab code in which I have 6 variables(=n_gc) with the min and max limit as "it.min_gc" and "it.max_gc" respectively. And each variable has 3(=M_g) values. From that I want to produce different combination i.e., as I have 729 combination (=3^6). So I used "ndgrid" operation and it produced the result which is stored in 6 variables(s1 u1 u2 ts vs vw) i.e., for example values stored in s1 is s1(1) to s1(729). But when I want to produce 729 combination again (i.e., when p=2) and try to store it, it shows the error "Assignment has more non-singleton rhs dimensions than non-singleton subscripts"

for k = 1:p
    for a = 1:n_gc
        gc(k,a,:) = (it.min_gc(k,a):((it.max_gc(k,a)-it.min_gc(k,a))/...
            (M_g-1)):it.max_gc(k,a));
    end
    [s1(k,:) u1(k,:) u2(k,:) ts(k,:) vs(k,:) vw(k,:)] = ndgrid(gc(k,1,:),...
        gc(k,2,:),gc(k,3,:),gc(k,4,:),gc(k,5,:),gc(k,6,:));

end

Please can anyone help me to resolve this issue?

Answer 1

[Walter Roberson]

1. it appears to me that you would be better rewriting your gc setting code in terms of linspace()
2. I think you might have to use squeeze() on each of the arguments to ndgrid(). This could, however, be avoided if you were to move k to the last subscript.
3. ndgrid is not going to be producing vectors, but you are trying to store the output as vectors
4. it might be easier to rewrite in terms of cell arrays, especially if you do not have a strong need to have individual names s1, u1, etc..

[Matt Learner]

Thank you. I used the linspace() for the gc setting code. I tried to use the squeeze() for each of the arguments to ndgrid(). It doesn't work. even tried to move k to the last subscript and again got the same error. Please can you explain more about your 4th comment about cell array? I used the ndgrid() to generate combinations of variables i.e., s1 u1 u2 ts vs and vw (each variable having 3 values equally spaced within the limits). The resulting grid for example s1(1) u1(1) u2(1) ts(1) vs(1) and vw(1) is the first grid combination and I will be using similar 729 grid points later for stage one (p=1). But for second stage p=2, I want to generate another set of 729 combination which is produced from the different regions of the 6 variables which I am not able to generate

[Walter Roberson]

ndgrid returns arrays with as many dimensions as there are number of variables. You cannot store those 6-dimensional arrays in a vector the way you are doing. You need to store them in an array, at least temporarily; once you have them temporarily stored you can reshape them in to vectors.

[s1t u1t u2t tst vst vwt] = ndgrid(gc(k,1,:), gc(k,2,:), gc(k,3,:), gc(k,4,:), gc(k,5,:), gc(k,6,:));

s1(k,:) = s1t(:); u1(k,:) = u1t(:); u2(k,:) = u2t(:);

ts(k,:) = tst(:); vs(k,:) = vst(:); vw(k,:) = vwt(:);

The squeeze() turns out not to be needed.

[Walter Roberson]

For using ndgrid with cells, see http://www.mathworks.com/matlabcentral/answers/27059-combinations-of-variables-and-step-sizing-for-creation-of-doe

[Matt Learner]

I already tried the same method of storing them in an array temporarily and tried to reshape it. But It gave me the error "??? Subscripted assignment dimension mismatch.. error in s1(k,:) = s1t(:);

[Matt Learner]

Please can you help me to resolve this issue?

[Walter Roberson]

For now, switch to

s1{k} = s1t(:); u1{k} = u1t(:); u2{k} = u2t(:);

ts{k} = tst(:); vs{k} = vst(:); vw{k} = vwt(:);

Then, after the k loop, look at the size() of each cell member of each of the variables, looking for one that is different. Then you should be able to track back to _why_ that one is different.

[Matt Learner]

I checked it too but s1 is not the only one. Reshaping is not working for any of the 6 variables i.e, obtained [s1t u1t u2t tst vst vwt] cannot be reshaped into vectors s1(k,:) = s1t(:); u1(k,:) = u1t(:); u2(k,:) = u2t(:);

ts(k,:) = tst(:); vs(k,:) = vst(:); vw(k,:) = vwt(:);

[Walter Roberson]

Right, so switch to cell arrays at the moment and look at the sizes of s1{1} vs s1{2} and so on.

[Matt Learner]

I switched to cell arrays. I am getting the following error message "??? Cell contents assignment to a non-cell array object." Also tried to store it into structure. Nothing is working fine for me. Please Help

[Walter Roberson]

Please show your current code.

[Matt Learner]

for k = 1:p

for a = 1:n_gc

gc(k,a,:) = linspace(it.min_gc(k,a),it.max_gc(k,a),M_g);

end

[s1t u1t u2t tst vst vwt] = ndgrid(gc(k,1,:), ...

gc(k,2,:),gc(k,3,:),gc(k,4,:),gc(k,5,:),gc(k,6,:));

s1{k} = s1t(:);

u1{k} = u1t(:);

u2{k} = u2t(:);

ts{k} = tst(:);

vs{k} = vst(:);

vw{k} = vwt(:);

end

[Matt Learner]

just a note, the inputs (p, n_gc,it.min_gc, it.max_gc, M_g) are working correctly

[Walter Roberson]

Are you running this as a script or in a function? If you were running it in a script, then because you would already have had the variables such as s1 and u1 in your workspace, you would have had trouble using those variables as cell arrays. You should use a function to protect against this kind of problem. Or at least at the top of your code, put in a "clear" statement

clear k a gc s1 u1 u2 ts vs vw

[Matt Learner]

Its a function, I am using the inputs (said in my last comment) to run this function.