How to index multiple matrices into one variable (or something comparable)

Hi,

I'm super new to Matlab. My first language is Python and this is my second but I'm still new to programming overall so please excuse my mistakes.

Basically I'm editing someone else's script for my own purposes. (The author of the script basically sent me an angry 'don't bother me' note so I can't ask him.)

His script contains a function which returns a matrix/array in a for loop, passing it to another for loop and recalculating the matrix for a particular "mode." I am separating that step from the proceeding step and trying to calculate all of the matrices and store them into one variable.

In python you can make lists of lists, i.e. y = [[1,2] , [2,4,5]] and y[1]=[2,4,5]. (I know the indexing init is different). Can you do the same in matlab? How would you do that?

Here is his original code:

    % mode defines the current eigenmode we want to visualize
    for mode = 1:numPCsToKeep
        eigenmode = eigenvecs(:,mode); % eigenmode direction
        variance = eigenvals(mode); % data variance in the current eigenmode
        standev = sqrt(variance);
        for i=-2:2
            % shift along the current eigenmode direction
            cfvec = mfvec + eigenmode'*i*standev;
            % visualize the current fvec
            cfvec = reshape(cfvec,3,256)';
            [spharm_vertices, spharm_faces] = surf_spharm(cfvec,dg,meshsize);
    %        subplot(4,5,(mode-1)*5+i+3); patch_lighta(spharm_vertices, spharm_faces); axis off;
     %       title(sprintf('%d*std',i));
            new_name = sprintf('PC%d_%d',mode,i);
            % make surface based on mesh picked
            switch outputFormat  % format to use for output
                case 'Amira'
                    outputToAmira(currentDir, new_name, spharm_vertices, spharm_faces);
                case 'STL'
                    outputToSTL(currentDir, new_name, spharm_vertices, spharm_faces);
                otherwise   % Amira output is default
                    outputToAmira(currentDir, new_name, spharm_vertices, spharm_faces);
            end
        end
    end

Here is my code:

    evcounter=1;
    eigenmode=[];
    variance=[];
    standev=[];
    for mode = 1:numPCsToKeep
        eigenmode(evcounter) = eigenvecs(:,mode); % eigenmode direction
        variance(evcounter) = eigenvals(mode); % data variance in the current eigenmode
        standev = sqrt(variance);
        evcounter = evcounter + 1;
    end
    % written for 3 PCs to be used even though numPCsToKeep allowed to be picked. Won't work if mode < 3 is to be used.

    for a=[-2 -1.5 -1 -0.5 0 0.5 1 1.5 2]    % PC1
      for b=[-2 -1.5 -1 -0.5 0 0.5 1 1.5 2]     % PC2
        for c=[-2 -1.5 -1 -0.5 0 0.5 1 1.5 2]     % PC3
                 % shift along the current eigenmode direction
                cfvec = mfvec + eigenmode(1)'*a*standev(1) + eigenmode(2)'*b*standdev(2) + eigenmode(3)'*c*standdev(3);
        % visualize the current fvec
                cfvec = reshape(cfvec,3,256)';
                [spharm_vertices, spharm_faces] = surf_spharm(cfvec,dg,meshsize);
      % subplot(4,5,(mode-1)*5+i+3); patch_lighta(spharm_vertices, spharm_faces); axis off;
       % title(sprintf('%d*std',i));
                new_name = sprintf('%dPC1_%dPC2_%PC3',a,b,c);
        % make surface based on mesh picked
                switch outputFormat  % format to use for output
                    case 'Amira'
                        outputToAmira(currentDir, new_name, spharm_vertices, spharm_faces);
                    case 'STL'
                        outputToSTL(currentDir, new_name, spharm_vertices, spharm_faces);
                    otherwise   % Amira output is default
                        outputToAmira(currentDir, new_name, spharm_vertices, spharm_faces);
            end
        end

Answer 1

[Walter Roberson]

If you want something list-like, use cell arrays. Your python

y = [[1,2] , [2,4,5]]

would translate to

y = {[1,2], [2,4,5]};

and then

y{1} would be [1,2] and y{2} would be [2,4,5]

Note that in Matlab, [2,4,5] is a numeric array of dimension 1 x 3, also known as a row vector. Numeric vectors are indexed with () subscripts, such as

t = y{2};
t(3)

with the t(3) being the third element of the numeric vector [2,4,5], which would be the scalar 5 . On the other hand, scalars in MATLAB are numeric arrays of dimension 1 x 1 -- the datatype for scalars and arrays is the same, so 5 is the same as [5]

This differs from t = {2,4,5} which will be a cell array of dimension 1 x 3. t(3) in this case (note the '()' indexing) would be the cell array {[5]} which is a cell array, not a scalar. t{3} would be the "content" of t(3) and would thus be [5], a scalar. Cell arrays may have cells that contain data of different size or data type than other cells: for example, z={'hello',[2,4,5]} is valid. z(1) would be {'hello'}, a cell containing a string; z{1} would be the string 'hello' itself; z(2) would be {[2,4,5]}, a cell containing a numeric array; z{2} would be [2,4,5], the numeric array itself. z{1}(2) would be 'e' and z{2}(2) would be 4. The distinction starts to matter more when you have more layers of cells -- cell arrays may contain cell arrays.

[Tina M.]

Thanks so much! I tweaked a bit more stuff too (see below) and it works!! Yay!

(For some reason it didn't understand how to transform an element of a cell matrix? I just made an extra step in my for loop where another temp variable is transformed and that worked like gangbusters.)

[Walter Roberson]

What kind of transformation would you like to do to the elements of the cell array? Generally speaking, cellfun() would be used, probably with the 'Uniform', 0 option added

Answer 2

[Tina M.]

I should also add that I'm trying to put these matrices into one item so that I can use a counter to call them in my for loop. I am also using nested for loops to generate all the points/models I'm creating because I don't know if there is some Matlab equivalent of itertools.