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

Introduction to deep reinforcement learning

Environment and neural network setup

DRL training loop

Introduction to deep Q learning

Deep learning and DQN

The Q-Network architecture

Instantiating the Q-Network

The barebone DQN algorithm

Barebone DQN action selection

Barebone DQN loss function

Training the barebone DQN

Introduction to Deep Reinforcement Learning

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.

DQN with experience replay

The double-ended queue

Experience replay buffer

The complete DQN algorithm

Epsilon-greediness

Fixed Q-targets

Implementing the complete DQN algorithm

Double DQN

Online network and target network in DDQN

Training the double DQN

Prioritized experience replay

Prioritized experience replay buffer

Sampling from the PER buffer

DQN with prioritized experience replay

Deep Q-learning

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.

Introduction to policy gradient

The policy network architecture

Working with discrete distributions

Policy gradient and REINFORCE

Action selection in REINFORCE

Training the REINFORCE algorithm

Advantage Actor Critic

Critic network

Actor Critic loss calculations

Training the A2C algorithm

Introduction to Policy Gradient Methods

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.

Proximal policy optimization

The clipped probability ratio

The clipped surrogate objective function

Entropy bonus and PPO

Entropy playground

Training the PPO algorithm

Batch updates in policy gradient

Minibatch and DRL

A2C with batch updates

Hyperparameter optimization with Optuna

Options

Hyperparameter

Not a hyperparameter

Hyperparameter or not?

Hands-on with Optuna

Congratulations!

Proximal Policy Optimization and DRL Tips

Embark on a journey to empower machines through Deep Reinforcement Learning (DRL). This course offers hands-on experience with powerful algorithms using PyTorch and Gymnasium.
Start with DRL foundations and traditional Reinforcement Learning, then implement Deep Q-Networks (DQN) with advanced refinements like Prioritized Experience Replay.
Advance your skills with policy-based methods and explore industry-standard algorithms like Proximal Policy Optimization (PPO) before optimizing your models using Optuna.

Discover the cutting-edge techniques that empower machines to learn and interact with their environments. You will dive into the world of Deep Reinforcement Learning (DRL) and gain hands-on experience with the most powerful algorithms driving the field forward. You will use PyTorch and the Gymnasium environment to build your own agents.

<h2>Master the Fundamentals of Deep Reinforcement Learning</h2>

Our journey begins with the foundations of DRL and their relationship to traditional Reinforcement Learning. From there, we swiftly move on to implementing Deep Q-Networks (DQN) in PyTorch, including advanced refinements such as Double DQN and Prioritized Experience Replay to supercharge your models.

Take your skills to the next level as you explore policy-based methods. You will learn and implement essential policy-gradient techniques such as REINFORCE and Actor-Critic methods.

<h2>Use Cutting-edge Algorithms</h2>

You will encounter powerful DRL algorithms commonly used in the industry today, including Proximal Policy Optimization (PPO). You will gain practical experience with the techniques driving breakthroughs in robotics, game AI, and beyond. Finally, you will learn to optimize your models using Optuna for hyperparameter tuning.

By the end of this course, you will have acquired the skills to apply these cutting-edge techniques to real-world problems and harness DRL's full potential!

Intermediate Deep Learning with PyTorch

Reinforcement Learning with Gymnasium in Python

Learn and use powerful Deep Reinforcement Learning algorithms, including refinement and optimization techniques.

Learn the Fundamentals of Deep Reinforcement Learning

Deep Reinforcement Learning in Python

Reinforcement Learning in Python

Minibatch and DRL

Deep Reinforcement Learning in Python

Latihan interaktif praktis