De geclipte surrogaatdoelfunctie

Implementeer de functie calculate_loss() voor PPO. Dit vraagt om het coderen van de kerninnovatie van PPO: de geclipte surrogaat-verliesfunctie. Deze helpt de beleidsupdate te begrenzen, zodat het beleid per stap niet te ver van het vorige beleid afwijkt.

De formule voor de geclipte surrogaatdoelfunctie is

In je omgeving staat de clipping-hyperparameter epsilon op 0,2.

Deze oefening maakt deel uit van de cursus

Deep Reinforcement Learning in Python

Cursus bekijken

Oefeninstructies

Bepaal de waarschijnlijkheidsratio's tussen \pi_\theta en \pi_{\theta_{old}} (ongeclipte en geclipte versies).
Bereken de surrogaatdoelen (ongeclipte en geclipte versies).
Bereken de geclipte surrogaatdoelfunctie van PPO.
Bereken het actorverlies.

Praktische interactieve oefening

Probeer deze oefening eens door deze voorbeeldcode in te vullen.

def calculate_losses(critic_network, action_log_prob, action_log_prob_old,
                     reward, state, next_state, done):
    value = critic_network(state)
    next_value = critic_network(next_state)
    td_target = (reward + gamma * next_value * (1-done))
    td_error = td_target - value
    # Obtain the probability ratios
    ____, ____ = calculate_ratios(action_log_prob, action_log_prob_old, epsilon=.2)
    # Calculate the surrogate objectives
    surr1 = ratio * ____.____()
    surr2 = clipped_ratio * ____.____()    
    # Calculate the clipped surrogate objective
    objective = torch.min(____, ____)
    # Calculate the actor loss
    actor_loss = ____
    critic_loss = td_error ** 2
    return actor_loss, critic_loss
  
actor_loss, critic_loss = calculate_losses(critic_network, action_log_prob, action_log_prob_old,
                                           reward, state, next_state, done)
print(actor_loss, critic_loss)

Code bewerken en uitvoeren

Deze oefening maakt deel uit van de cursus

Deep Reinforcement Learning in Python

SkillTag.level.advancedSkillTag.label

4.8+

Begin de cursus gratis

Discover how deep reinforcement learning improves upon traditional Reinforcement Learning while studying and implementing your first Deep Q Learning algorithm.

Exercise 1: Introduction to deep reinforcement learning Exercise 2: Environment and neural network setup Exercise 3: DRL training loop Exercise 4: Introduction to deep Q learning Exercise 5: Deep learning and DQN Exercise 6: The Q-Network architecture Exercise 7: Instantiating the Q-Network Exercise 8: The barebone DQN algorithm Exercise 9: Barebone DQN action selection Exercise 10: Barebone DQN loss function Exercise 11: Training the barebone DQN

Dive into Deep Q-learning by implementing the original DQN algorithm, featuring Experience Replay, epsilon-greediness and fixed Q-targets. Beyond DQN, you will then explore two fascinating extensions that improve the performance and stability of Deep Q-learning: Double DQN and Prioritized Experience Replay.

Exercise 1: DQN with experience replay Exercise 2: The double-ended queue Exercise 3: Experience replay buffer Exercise 4: DQN with experience replay Exercise 5: The complete DQN algorithm Exercise 6: Epsilon-greediness Exercise 7: Fixed Q-targets Exercise 8: Implementing the complete DQN algorithm Exercise 9: Double DQN Exercise 10: Online network and target network in DDQN Exercise 11: Training the double DQN Exercise 12: Prioritized experience replay Exercise 13: Prioritized experience replay buffer Exercise 14: Sampling from the PER buffer Exercise 15: DQN with prioritized experience replay

Learn about the foundational concepts of policy gradient methods found in DRL. You will begin with the policy gradient theorem, which forms the basis for these methods. Then, you will implement the REINFORCE algorithm, a powerful approach to learning policies. The chapter will then guide you through Actor-Critic methods, focusing on the Advantage Actor-Critic (A2C) algorithm, which combines the strengths of both policy gradient and value-based methods to enhance learning efficiency and stability.

Exercise 1: Introduction to policy gradient Exercise 2: The policy network architecture Exercise 3: Working with discrete distributions Exercise 4: Policy gradient and REINFORCE Exercise 5: Action selection in REINFORCE Exercise 6: Training the REINFORCE algorithm Exercise 7: Advantage Actor Critic Exercise 8: Critic network Exercise 9: Actor Critic loss calculations Exercise 10: Training the A2C algorithm

Explore Proximal Policy Optimization (PPO) for robust DRL performance. Next, you will examine using an entropy bonus in PPO, which encourages exploration by preventing premature convergence to deterministic policies. You'll also learn about batch updates in policy gradient methods. Finally, you will learn about hyperparameter optimization with Optuna, a powerful tool for optimizing performance in your DRL models.

Exercise 1: Proximal policy optimization Exercise 2: De afgeknipte waarschijnlijkheidsratio Exercise 3: De geclipte surrogaatdoelfunctie

Huidige oefening

Exercise 4: Entropiebonus en PPO Exercise 5: Entropie-speeltuin Exercise 6: De PPO-algoritme trainen Exercise 7: Batch-updates in policy gradient Exercise 8: Minibatch en DRL Exercise 9: A2C met batch-updates Exercise 10: Hyperparameteroptimalisatie met Optuna Exercise 11: Hyperparameter of niet?Exercise 12: Aan de slag met Optuna Exercise 13: Gefeliciteerd!