Transform algorithm avoiding loops

Hi, I have to transform the following algorithm in an algorithm without loops.

Consider a matrix A with the following characteristics:

A=[[1 2; 1 3; 1 4; 2 1; 2 6; 3 1; 3 2; 3 3; 3 4; 3 5] randi([0,1], 10,2)];

In particular, the characteristics of A which I can rely on are: the first column shows all increasing numbers from 1 to n=3, but we don't know how many times each number is repeated; for each number from 1 to n=3, some numbers between 1 and m=6 are associated in the second column, not necessarily in an increasing order.

I construct a vector Y of dimension (size(A,1))x1 in the following way:

   n=max(A(:,1));
   m=max(A(:,2));
   gamma=2;
   delta=-3;
   Y=zeros(size(A,1),2);
   for t=1:m
   At=A(A(:,2)==t,:);
   for l=1:size(At,1)
   Yl= mvnrnd((gamma/delta*At(l,3:end))',eye(n-1)); %1x(n-1)
   ismember(A(:,1:2),[At(l,1) t],'rows');
   Y(index,:)=Yl
   end
   end

I want to rewrite the algorithm above without loops. Could you help me?