Double complex-Error using quadprog

Hi, I am converting data type of double since it has a big number to a normalize form. When i used the normalized data for SVM Classification, it turns out the accuracy is very low. So I decided to use log base 2 and then I put on SVM Classification I've got these answer

Error using quadprog (line 318) All inputs to QUADPROG must be real.

Error in svmtrain (line 528) [alpha, ~, exitflag, output] = quadprog(H,-ones(nPoints,1),[],[],...

How do I solve this?

Answer 1

[Walter Roberson]

Whatever values you are taking log2() of, appears to have some negative values, so the log2 is imaginary.