How can I re-create sparse matrixes from the row and column indexes, and values?

Hi there,

I have a struct Z, which contain 28 structs(Z1,....,Z28).

Each of them has

i (int32) = whose numbers represent the row indexes

j (int32) = whose numbers represent the columns indexes

x (double) = with values

This data was imported from 28 sparses matrixes (24000x24000), using R, in a way that the length of i,j and x are the same, but x(1,1) does not necessarily populate Z(i1,j1). For example, considering Z1, the first value of x is actually the element z(126,126).

Now, in Matlab, I need to recreate these 28 matrixes (24000x24000) matching the values of x with the i,j indexes.

Any clue on how to solve it. I would be extremely grateful!

Using a loop would be the best!

Cheers,

[Tiago Furlanetto]

Thanks foryour message, Torsten.

Z1,..., Z28 are the tow-dimensional matrixes I inported from R.

with z(126,126), I meant that first valeu of x is located in row 126, column 126 of matrix Z1.

Did I explain myself?

Answer 1

[Steven Lord]

Let's take a simple example:

rng default
S = sprand(10, 10, 0.1); % 10-by-10 with roughly 10 nonzero elements

Let's get the vectors of row and column indices and values

[rows, cols, values] = find(S);

Recreate S using rows, cols, values. In this case S has elements in both the last row and last column but to handle the general case where it didn't let's also state how large it needs to be.

S2 = sparse(rows, cols, values, 10, 10);

Is S the same as S2? Yes.

isequal(S, S2)

[Tiago Furlanetto]

Steven,

thanks for your message.

I have tried as you suggested and some with some variations on the code,

however, once Z is a struct, as well as Z1,....Z28. I was not able to run the operation "find"

In the figure bellow are the variables (i,j,x) from each Z struct.

Could you help me understand how to call this variables to perform the "find" operation?

[Tiago Furlanetto]

Steven,

figured it out.

Z_cell = struct2cell(Z);

Z_2013 = Z_cell{28};

%%

Z1_cell = struct2cell(Z1);

r = Z1_cell{1};

r = double(r);

c = Z1_cell{2};

c = double(c);

v = Z1_cell{3};

%%

S2 = sparse(r, c, v, 24000, 24000);

Thanks a lot

[Steven Lord]

If you already have the vectors of indices and the vector of values, you don't need to call find. I didn't have them (I had the original sparse matrix) so I used find to generate those vectors. Start with the next step.

[Tiago Furlanetto]

Sure! make sense.

Many thanks.