Problem with symsum function: doesn't work with matrix elements?

I am trying to sum elements of a matrix with the symsum function.

E = squareform(pdist(ConcernedNodes2(:,1:2)));
DistMin = min(DistRoadMetro,[],3);
N = length(ConcernedNodes2);
V = zeros(length(ConcernedNodes2),timeSteps);

for t = 1:timeSteps
    C = perc_road_num(:,:,t);
    for i=1:length(ConcernedNodes2)
        V(i,t) = (1/(symsum(E(i,k),k,1,length(ConcernedNodes2))*(N-1)))*symsum(E(i,j).*C(i,j)/(DistMin(i,j)),j,1,length(ConcernedNodes2));
    end
end

And I get this error message: Undefined function 'symsum' for input arguments of type 'double'.

Can someone help me please?

[madhan ravi]

Perhaps you want to use cumsum().

[Habib]

Nope, I need the sum of all elements of line i of the matrices in the equation.

[madhan ravi]

Then why not just use sum() since it's one dimensional inside the loop?

[Habib]

for t = 1:timeSteps
    C = perc_road_num(:,:,t);
    for i=1:length(ConcernedNodes2)
        V(i,t) = (1./sum(E(i,1:length(ConcernedNodes2))*(N-1))).*sum(E(i,1:length(ConcernedNodes2)).*C(i,1:length(ConcernedNodes2))./(DistMin(i,1:length(ConcernedNodes2))));

    end
end

I tried this, but still doesn't work.

It runs but the results are NaN

[Walter Roberson]

I suspect some of those sum come out as 0

[Habib]

for t = 1:timeSteps
    C = perc_road_num(:,:,t);
    for i=1:length(ConcernedNodes2)
        V2(i,t) = (1./sum(E(i,1:length(ConcernedNodes2)~=i)*(N-1))).*sum(E(i,1:length(ConcernedNodes2)~=i).*C(i,1:length(ConcernedNodes2)~=i)./(DistMin(i,1:length(ConcernedNodes2)~=i)));
    end
end

This works, I forgot to add ~=i in the sum

Answer 1

[Adam Danz]

The first two inputs to symsum() should be symbolic. It's difficult to interpret what the code is supposed to do without knowing what's stored in the variables.

More info in symsum inputs:

https://www.mathworks.com/help/symbolic/symsum.html#btydjwt-2

Examples:

https://www.mathworks.com/help/symbolic/symsum.html#btydjwt-4

It seems like you don't need symsum() in the first place. In this segment below, what I think you're trying to do is sum part of a row of matrix 'E'. It looks like you're trying to take the sum of row 'i' from column 1 to column length(ConcernedNodes2).

symsum(E(i,k),k,1,length(ConcernedNodes2)) % <--- incorrect 

If that's the case, what you really want is

sum(E(i,1:length(ConcernedNodes2)))   % <--- correct

which will give you one value (the sum of those numbers). Or, perhaps you are looking for cumsum() which would be cumulative sum and would output a vector the same length as the segment you're isolating.

[Habib]

Sorry, E is a matrix

[Habib]

So which function should I use?

[Adam Danz]

I've updated my answer with suggestions.

[Habib]

Actually this worked well:

for t = 1:timeSteps
    C = perc_road_num(:,:,t);
    for i=1:length(ConcernedNodes2)
        U1=0;
        U2=0;
        for j = 1:length(ConcernedNodes2)
            if j ~=i
                U1 = U1 + E(i,j)*C(i,j)/(DistMin(i,j));
                U2 = U2 + E(i,j);
            end           
        end
        V(i,t) = 1/(U2*(N-1))*U1;
    end
end

And it is quicker than using the sum function :)