How to add elements to a matrix one-by-one?

I have a matrix [0 1; 1 0]. I need to:

1) choose random row from this matrix

2) duplicate every non-zero element from this row with a some probability and add it to the matrix.

3) enlarge the size of the matrix (to keep it square)

4) repeate the procedure with the new matrix (with added row), say, 10 times.

The idea is to use this matrix to construct a graph (where zeros correspond to the absence of linkadge between two nodes and ones - to the presence of linkadge).

Now I stuck on the stage of enlarging the matrix: as every element of the randomly chosen row should be checked alone, matlab do not allow to add elements one-by-one in one cycle.

conmat = [0 1; 1 0] ;
S=size(conmat,1);
r=randi(S);
sigma=0.3;
 a=rand;
rndrow=conmat(r,:);
rowadd=[];
for j=1:size(conmat,1);
       if sigma>a
        rowadd(j)=[rowadd rndrow(j)]
    elseif sigma<a
        rowadd(j)= [0]
    end
conmat=[conmat rowadd] %ERROR
coladd=transpose(conmat(3,:)) 
coladd=[coladd; 0]
conmat=[conmat coladd]
end

Error using horzcat
Dimensions of arrays being concatenated are not consistent.

Error in network_181129 (line 15)
conmat=[conmat rowadd]

I feel I need to use some different approach not syntaxis tricks but I'm not so experinced in coding and Matlab and already spent a lot of time on this problem. I will appreciate any help and ideas, thank you in advance!

Answer 1

[Luna]

Hi,

I tried to fix the errors (cut and paste some lines inside the inner for loop), and added outer for loop for 10 times calculation(I assume your random variable a changes in every iteration).

Try this below. It takes a little bit long time since I ended up with conmat 2048x2048 matrix. But maybe there could be a high performanced version of it.

conmat = [0 1; 1 0] ;
for m=1:10
    S=size(conmat,1);
    r=randi(S);
    sigma=0.3;
    a=rand;
      
    for j=1:size(conmat,1)
        rowadd=zeros(1,size(conmat,1));
        rndrow=conmat(r,:);
        if sigma > a
            rowadd= rndrow;
        elseif sigma < a
            %         rowadd(j)= 0
        end
        conmat=[conmat; rowadd]; %ERROR
        coladd=transpose(conmat(3,:));
        coladd=[coladd; 0];
        conmat=[conmat coladd];
    end   
end

If you want to replicate matrices you can use repmat function also. Here the link: repmat

[Elizaveta Grigorashvili]

Hi Luna, thank you a lot! It was reaaly helpful. I changed a little the code you've suggested and obtained the needed graph. Now the script looks like that:

conmat = [0 1; 1 0] ;

for m=1:10
    S=size(conmat,1);
    r=randi(S);
    a=rand;
    sigma=0.5;
    for j=1:size(conmat,1)
        rowadd=zeros(1,size(conmat,1));
        rndrow=conmat(r,:);
        if sigma > a
            rowadd(j)= rndrow(j);
            rowadd(j+1)=1;
        elseif sigma < a
            rowadd(j)= rndrow(j);
            rowadd(j+1)==0;
        end
        conmat=[conmat; rowadd]; 
        coladd=transpose(rowadd);
        coladd=[coladd; 0];
        conmat=[conmat coladd];
        conmat( ~any(conmat,2), : ) = [];
        conmat( :, ~any(conmat,1) ) = []; 
    end
    
end
G=graph(conmat);
plot(G)

[Luna]

Glad to be helpful! :)