Sampled-data Control of Probabilistic Boolean Control Networks: A Deep Reinforcement Learning Approach