'Simplification code to avoid if else division problem

I would like to make an optimisation by avoiding to use theIf else i used this element because when E = 0in the logic of my probelm B = [0 0 0 0 0].

By consequence I have an error Nan . I suppose it is because i have R/0 at i =1 and 2

% i reflects the time
% boolean flag type double
B = [0 0 0 0 0
     0 0 0 0 0
     0 0 0 0 1
     0 0 0 1 0
     0 0 0 1 1];
% info1
E = [0 0 10 20 40];

% info2
R(1:5) = 1/30;

powerload_R2 = zeros(5);

for i =1:5
    if E(i)>0
       powerload_R(i,:) = R ./ dot(R,B(i,:)) .* B(i,:)*E(i); % fonctionnel
    else
       powerload_R(i,:) = 0;
    end
    powerload_R2(i,:) = R ./ dot(R,B(i,:)) .* B(i,:)*E(i); %
end
%results

%what  i get with :  
powerload_R(i,:)=

0   0   0   0   0
0   0   0   0   0
0   0   0   0   10
0   0   0   20  0
0   0   0   20  20

powerload_R2(i,:)=
NaN NaN NaN NaN NaN
NaN NaN NaN NaN NaN
0   0   0   0   10
0   0   0   20  0
0   0   0   20  20
matlabmath


Solution 1:[1]

The NaNs appear from 0/0 division, so only add eps to the denominator to avoid this and get 0/eps = 0 instead.

% i reflects the time
% boolean flag type double
B = [0 0 0 0 0
     0 0 0 0 0
     0 0 0 0 1
     0 0 0 1 0
     0 0 0 1 1];
% info1
E = [0 0 10 20 40];

% info2
R(1:5) = 1/30;

powerload_R  = zeros(5);
powerload_R2 = zeros(5);

for i = 1:5 
    if E(i) > 0
        powerload_R(i,:) = R ./ dot(R,B(i,:)) .* B(i,:)*E(i); % fonctionnel
    end
    powerload_R2(i,:) = R ./ (dot(R,B(i,:))+eps) .* B(i,:)*E(i); %
end

Solution 2:[2]

Just for fun, the vectorized version of your formula is:

powerload_R = R ./ (B*R.') .* (B.*E.');

This results in:

powerload_R =
   NaN   NaN   NaN   NaN   NaN
   NaN   NaN   NaN   NaN   NaN
     0     0     0     0    10
     0     0     0    20     0
     0     0     0    20    20

We can make a logical array to tell calculate only the rows where E is non-zero.

E_NZ = E ~= 0;

Then use E_NZ to index B and E:

powerload_R(E_NZ,:) = R ./ (B(E_NZ,:)*R.') .* (B(E_NZ,:).*E(E_NZ).');

So the complete script would look like this:

% i reflects the time
% boolean flag type double
B = [0 0 0 0 0
     0 0 0 0 0
     0 0 0 0 1
     0 0 0 1 0
     0 0 0 1 1];
% info1
E = [0 0 10 20 40];

% info2
R(1:5) = 1/30;

powerload_R  = zeros(5);
E_NZ = E ~= 0;

powerload_R(E_NZ,:) = R ./ (B(E_NZ,:)*R.') .* (B(E_NZ,:).*E(E_NZ).');

The result is the same matrix without the NaN values and no (explicit) looping:

powerload_R =
    0    0    0    0    0
    0    0    0    0    0
    0    0    0    0   10
    0    0    0   20    0
    0    0    0   20   20

Sources

This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.

Source: Stack Overflow

Solution Source
Solution 1 AboAmmar
Solution 2

Related Questions