Error using custom RBF Kernel

I am using a rbf kernel:

    function G = myrbfgauss(U,V)
    sigma=1.5;
    Xny = U-V;
    Normxny = Xny'*Xny;
    G = exp(-Normxny/(2*sigma^3));
    end

With the Matlab example code found on the Machine Learning/Concepts/SVM help file under "Train SVM Classifiers Using a Custom Kernel". I replace the example's sigmoid kernel with the rbf code above and get the following error:

Error using solve Function myrbfgauss errors with the following message: Matrix dimensions must agree.

Can someone please help?