I have an n-by-m matrix that I want to convert to a mn-by-m matrix, with each m-by-m block of the result containing the diagonal of each row.
For example, if the input is:
[1 2; 3 4; 5 6]
the output should be:
[1 0; 0 2; 3 0; 0 4; 5 0; 0 6]
Of course, I don't want to assemble the matrix step by step myself with a for loop.
Is there a vectorized and simple way to achieve this?
For a vectorized way to do this, create the linear indices of the diagonal elements into the resulting matrix, and assign directly.
%# create some input data
inArray = [10 11;12 13;14 15];
%# make the index array
[nr,nc]=size(inArray);
idxArray = reshape(1:nr*nc,nc,nr)';
idxArray = bsxfun(#plus,idxArray,0:nr*nc:nr*nc^2-1);
%# create output
out = zeros(nr*nc,nc);
out(idxArray) = inArray(:);
out =
10 0
0 11
12 0
0 13
14 0
0 15
Here's a simple vectorized solution, assuming X is the input matrix:
Y = repmat(eye(size(X, 2)), size(X, 1), 1);
Y(find(Y)) = X;
Another alternative is to use sparse, and this can be written as a neat one-liner:
Y = full(sparse(1:numel(X), repmat(1:size(X, 2), 1, size(X, 1)), X'));
The easiest way I see to do this is actually quite simple, using simple index referencing and the reshape function:
I = [1 2; 3 4; 5 6];
J(:,[1,4]) = I;
K = reshape(J',2,6)';
If you examine J, it looks like this:
J =
1 0 0 2
3 0 0 4
5 0 0 6
Matrix K is just what wanted:
K =
1 0
0 2
3 0
0 4
5 0
0 6
As Eitan T has noted in the comments, the above is specific to the example, and doesn't cover the general solution. So below is the general solution, with m and n as described in the question.
J(:,1:(m+1):m^2) = I;
K=reshape(J',m,m*n)';
If you want to test it to see it working, just use
I=reshape(1:(m*n),m,n)';
Note: if J already exists, this can cause problems. In this case, you need to also use
J=zeros(n,m^2);
It may not be the most computationally efficient solution, but here's a 1-liner using kron:
A = [1 2; 3 4; 5 6];
B = diag(reshape(A', 6, 1) * kron(ones(3, 1), eye(2))
% B =
% 1 0
% 0 2
% 3 0
% 0 4
% 5 0
% 0 6
This can be generalized if A is n x m:
diag(reshape(A.', n*m, 1)) * kron(ones(n,1), eye(m))
Related
I have a vector that contains 5 numbers and I want to pad it with zeros. How can I do it?
A = [1 2 3 4 5].';
I want the zero padded vector to be like this:
A_new = [0 0 0 0 0 1 2 3 4 5].';
Also, for another case, I want to assign 1, 3, 4 to matrix W as follows, with all else being zeros. The length of W is 7. W = [0 1 0 0 3 0 4].
You can use following code
newA = [zeros(5,1); A]
About another case. You need something like
inds = [2 5 7];
elems = [1 3 4];
W = zeros(7,1);
W(inds) = elems
Given this vector
a = [1 2 3 4]
I want to create a matrix like this
b = [1 0 0 0;
2 1 0 0;
3 2 1 0;
4 3 2 1;
0 4 3 2;
0 0 4 3;
0 0 0 4]
in a vectorized way not using loops.
Hint: use conv2 (hover mouse to see code):
a = [1 2 3 4];
b = conv2(a(:), eye(numel(a)));
Or, in a similar mood, you can use convmtx (from the Signal Processing Toolbox):
a = [1 2 3 4];
b = convmtx(a(:), numel(a));
One way to do it:
a = [1 2 3 4]
n = numel(a);
%// create circulant matrix from input vector
b = gallery('circul',[a zeros(1,n-1)]).' %'
%// crop the result
c = b(:,1:n)
Another way:
b = union( tril(toeplitz(a)), triu(toeplitz(fliplr(a))),'rows','stable')
or its slightly variation
b = union( toeplitz(a,a.*0),toeplitz(fliplr(a),a.*0).','rows','stable')
and probably even faster:
b = [ toeplitz(a,a.*0) ; toeplitz(fliplr(a),a.*0).' ]
b(numel(a),:) = []
With bsxfun -
na = numel(a)
b = zeros(2*na-1,na)
b(bsxfun(#plus,[1:na]',[0:na-1]*2*na)) = repmat(a(:),1,na)
If you are looking for a faster pre-allocation, you can do -
b(2*na-1,na) = 0;.
Another bsxfun -
a=[1 2 3 4];
m=numel(a);
b=[a,zeros(1,m-1)].';
Q=bsxfun(#circshift, b, [0:m-1])
I have one matrix like below-
A=[1 1 1 1 1;
0 1 1 1 2;
0 0 1 1 3]
But I want to place all the 0 at the end of the row, so A should be like-
A=[1 1 1 1 1;
1 1 1 2 0;
1 1 3 0 0]
How can I do this? Matlab experts please help me.
There you go. Whole matrix, no loops, works even for non-contiguous zeros:
A = [1 1 1 1 1; 0 1 1 1 2; 0 0 1 1 3];
At = A.'; %// It's easier to work with the transpose
[~, rows] = sort(At~=0,'descend'); %// This is the important part.
%// It sends the zeros to the end of each column
cols = repmat(1:size(At,2),size(At,1),1);
ind = sub2ind(size(At),rows(:),cols(:));
sol = repmat(NaN,size(At,1),size(At,2));
sol(:) = At(ind);
sol = sol.'; %'// undo transpose
As usual, for Matlab versions that do not support the ~ symbol on function return, change ~ by a dummy variable, for example:
[nada, rows] = sort(At~=0,'descend'); %// This is the important part.
A more generic example:
A = [1 3 0 1 1;
0 1 1 1 2;
0 0 1 1 3]
% Sort columns directly
[~,srtcol] = sort(A == 0,2);
% Sorted positions
sz = size(A);
pos = bsxfun(#plus, (srtcol-1)*sz(1), (1:sz(1))'); % or use sub2ind
The result
B = A(pos)
B =
1 3 1 1 0
1 1 1 2 0
1 1 3 0 0
there are many ways to do this. one fast way can be easily like this:
a = [1 2 3 4 0 5 7 0];
idx=(find(a==0));
idx =
5 8
b=a; % save a new copy of the vector
b(idx)=[]; % remove zero elements
b =
1 2 3 4 5 7
c=[b zeros(size(idx))]
c =
1 2 3 4 5 7 0 0
You may modify this code as well.
If your zeros are always together, you could use the circshift command. This shifts values in an array by a specified number of places, and wraps values that run off the edge over to the other side. It looks like you would need to do this separately for each row in A, so in your example above you could try:
A(2,:) = circshift(A(2,:), [1 -1]); % shift the second row one to the left with wrapping
A(3,:) = circshift(A(3,:), [1 -2]); % shift the third row two to the left with wrapping
In general, if your zeros are always at the front of the row in A, you could try something like:
for ii = 1:size(A,1) % iterate over rows in A
numShift = numel(find(A(ii,:) == 0)); % assuming zeros at the front of the row, this is how many times we have to shift the row.
A(ii,:) = circshift(A(ii,:), [1 -numShift]); % shift it
end
Try this (just a fast hack):
for row_k = 1:size(A, 1)
[A_sorted, A_sortmap] = sort(A(row_k, :) == 0, 'ascend');
% update row in A:
A(row_k, :) = A(row_k, A_sortmap);
end
Now optimized for versions of MATLAB not supporting ~ as garbage lhs-identifier.
#LuisMendo's answer is inspiring in its elegance, but I couldn't get it to work (perhaps a matlab version thing). The following (based on his answer) worked for me:
Aaux = fliplr(reshape([1:numel(A)],size(A)));
Aaux(find(A==0))=0;
[Asort iso]=sort(Aaux.',1,'descend');
iso = iso + repmat([0:size(A,1)-1]*size(A,2),size(A,2),1);
A=A.';
A(iso).'
I've also asked this question and got a super elegant answer (non of above answers is same) here:
Optimize deleting matrix leading zeros in MATLAB
Suppose I have a matrix A, and I'd like to get the matrix [A 0; 0 1]. Is there a built function to do this?
So if my matrix is [2 3; 1 4], I'd get back [2 3 0; 1 4 0; 0 0 1]
The easiest way is:
newA = A;
newA(end+1,end+1) = 1;
This works because you can index outside an array for assignments, because end indicates the last element (here in row and column), and because Matlab pads with zeros when you grow an array. If you just want to grow A, you can even skip the creation of newA, of course.
I always use matrix concatenation for problems like this
So for your example:
A = [2 3; 1 4]
A = [A A(:,1)*0; A(1,:)*0 1]
produces
A =
2 3 0
1 4 0
0 0 1
The nice thing about this trick is that its very flexible and you can do all sorts of tranformations
very easily. For example
A = [2 3; 1 4]
A = [1 A(1,:)*0; A(:,1)*0 A]
produces
A =
1 0 0
0 2 3
0 1 4
I've got a matrix A with the dimensions m X n. For every column i (i > 0and i <= n) I want to flip a coin and fill the whole column with 0 values with probability p. How can this be accomplished in MATLAB?
Example:
A = [1 2 3 4; 5 6 7 8] and p = 0.5 could result in
A' = [1 0 3 0; 5 0 7 0]
You can use the function rand() to generate an array of uniformly distributed random numbers, and use logical indexing to select colums where that array is less than p:
A = [1 2 3 4; 5 6 7 8];
p = 0.5;
A(:, rand(size(A,2), 1)<p) = 0
A =
0 2 0 0
0 6 0 0
You can do something like bsxfun(#times, A, rand(1, size(A, 2)) > p). Alex's answer is admittedly better, though.