StabilizingDynamicalSystemsviaPolicyGradient Methods