Gradient Descent produces incorrect Thetas in octave

Question

I\'m trying out a prediction algorithm using polynomial regression of the form h(x) = theta0 + theta1 * x1 + theta2 * x2, where x2=x1^2