Initializing the deep Q-learning agent

Now, we have programmatically defined all of the individual components that are necessary to initialize our deep Q-learning agent. For this, we use the imported DQNAgent object from rl.agents.dqn and defined the appropriate parameters, as shown here:

#Initialize the atari_processor() class

processor = AtariProcessor()

# Initialize the DQN agent 
dqn = DQNAgent(model=model,             #Compiled neural network model
               nb_actions=nb_actions,   #Action space
               policy=policy,   #Policy chosen (Try Boltzman Q policy)
               memory=memory,   #Replay memory (Try Episode Parameter  
                                                memory)
               processor=processor,     #Atari processor class
#Warmup steps to ignore initially (due to random initial weights)
               nb_steps_warmup=50000,   
       ...