Basic Vectorization Question (speed required)

I have a 2D matrix of data. I have an index to reduce that 2D matrix to a 1D matrix.

How can I do it without using a for loop? Ie by vectorization as it needs to be super quick.

thanks

Z = rand(5,100); %matrix of data
IDY = randi([1 5], [1,100]); %index into that matrix
X = NaN(size(Z,2),1); %pre allocate
X = Z(IDY,:); %this fails as it does not generate a 1D matrix (vector) as required

this is how I can do it using a loop

X2 = NaN(size(Z,2),1);
for i = 1: size(X,1)
    X2(i) = Z(IDY(i),i);
end

Answer 1

[Azzi Abdelmalek]

Z = rand(5,100); %matrix of data
IDY = randi([1 5], [1,100]); %index into that matrix
X = NaN(size(Z,2),1); %pre allocate
X = Z(IDY,:)
X3=Z(sub2ind(size(Z),IDY,1:size(Z,2)))

[Matlab2010]

thats a great answer. thank you!

PS have never used sub2ind before but will read up on it now

Answer 2

[Image Analyst]

To get all of Z in row-major order, do:

X = Z(:);

To get random ordering, like your code attempted to do, try this:

Z = rand(5,100) %matrix of data
randomIndexes = randperm(numel(Z)) %index into that matrix
X = Z(randomIndexes)

[Matlab2010]

I must have explained my question very badly.

I have an index of size 1*T (in my example T = 100). Each value in the index can take a value 1: K (K = 5) in my example. This is an index value for each row in the K*T data matrix. I wish to apply the index to generate a matrix 1*T in size.

Your example code generates an index of values 1: K*T and then produces an a matrix of K*T in size.

thanks