Currently I am using RL agent DQN to predict action and update the action value function. But if I have a constraint to run a particular action for n times, can I have a constraint in DQN specified for the agent to take action? If yes, How can I possibly do it (using TF)?
Thank you in advance