Convergence of the conjugate gradient method with computationally convenient modifications