On The Application Of Continuous Deterministic Reinforcement Learning In Neural Architecture Search