warning: Matrix is singular

Hi,

When i run the code, i get this warning

Matrix is singular, close to singular or badly scaled. Results may be   inaccurate.
RCOND = NaN.

and the rest matrices v,var have inf values.

The matrix where i take this warning is 9x9. I attach part of the code to see better where is my fault.

for i=1:num_exp,

          distance_hat=distance.*10.^(0.05*sqrt(sigma)*randn(1,N));
          m = 2.3026*log10(distance_hat);
          var = exp(4*m)*(exp(8*sigma.^2)-exp(4*sigma.^2));
          v = var(1,1)+var(1,2:N);
          S = eye(N-1).*diag(v) + ~eye(N-1).*var(1,1)
          S12 = inv(S)
          sol = loc(x,distance_hat,S12);
          weighted = loc(x,distance_hat,S12);
  end

what should i change to not take this warning?

[Walter Roberson]

You never use "i" in the loop, and all of the variables you create get overwritten on the next iteration of the "for i" loop so the result is the same as if you had only done the final iteration of the loop.

This suggests that either you have mis-written your code or else that you have not shown us the actual code.

[Dimitris Georgis]

I use "i" for experiment loop and find the average error. So the rest command to end the for loop are:

error(i)=sqrt((sol(1)-u(1))^2+(sol(2)-u(2))^2);

[Jan]

Do not overwrite the important function "error" by a variable.

Answer 1

[Jan]

I guess that the line, which causes the error, is:

S12 = inv(S)

(Please post the complete error message - most of all it is inefficient to let the readers guess, where the error appears.) Then the message means, that there is no inverse of S. Because we do not know, why you create S in the shown way, there is no chance to suggest an improvement.

[Dimitris Georgis]

The complete error message is: "In least_squares_loc (line 15) In run_loc (line 31) Warning: Matrix is singular, close to singular or badly scaled. Results may be inaccurate. RCOND = NaN."

I have attach the code to run it, and check.