4.41 out of 5
4.41
193 reviews on Udemy

Modern Reinforcement Learning: Actor-Critic Algorithms

How to Implement Cutting Edge Artificial Intelligence Research Papers in the Open AI Gym Using the PyTorch Framework
Instructor:
Phil Tabor
1,149 students enrolled
English [Auto]
How to code policy gradient methods in PyTorch
How to code Deep Deterministic Policy Gradients (DDPG) in PyTorch
How to code Twin Delayed Deep Deterministic Policy Gradients (TD3) in PyTorch
How to code actor critic algorithms in PyTorch
How to implement cutting edge artificial intelligence research papers in Python

In this advanced course on deep reinforcement learning, you will learn how to implement policy gradient, actor critic, deep deterministic policy gradient (DDPG), twin delayed deep deterministic policy gradient (TD3), and soft actor critic (SAC) algorithms in a variety of challenging environments from the Open AI gym. There will be a strong focus on dealing with environments with continuous action spaces, which is of particular interest for those looking to do research into robotic control with deep reinforcement learning.

Rather than being a course that spoon feeds the student, here you are going to learn to read deep reinforcement learning research papers on your own, and implement them from scratch. You will learn a repeatable framework for quickly implementing the algorithms in advanced research papers. Mastering the content in this course will be a quantum leap in your capabilities as an artificial intelligence engineer, and will put you in a league of your own among students who are reliant on others to break down complex ideas for them.

Fear not, if it’s been a while since your last reinforcement learning course, we will begin with a briskly paced review of core topics.

The course begins with a practical review of the fundamentals of reinforcement learning, including topics such as:

  • The Bellman Equation

  • Markov Decision Processes

  • Monte Carlo Prediction

  • Monte Carlo Control

  • Temporal Difference Prediction TD(0)

  • Temporal Difference Control with Q Learning

And moves straight into coding up our first agent: a blackjack playing artificial intelligence. From there we will progress to teaching an agent to balance the cart pole using Q learning.

After mastering the fundamentals, the pace quickens, and we move straight into an introduction to policy gradient methods. We cover the REINFORCE algorithm, and use it to teach an artificial intelligence to land on the moon in the lunar lander environment from the Open AI gym. Next we progress to coding up the one step actor critic algorithm, to again beat the lunar lander.

With the fundamentals out of the way, we move on to our harder projects: implementing deep reinforcement learning research papers. We will start with Deep Deterministic Policy Gradients (DDPG), which is an algorithm for teaching robots to excel at a variety of continuous control tasks. DDPG combines many of the advances of Deep Q Learning with traditional actor critic methods to achieve state of the art results in environments with continuous action spaces.

Next, we implement a state of the art artificial intelligence algorithm: Twin Delayed Deep Deterministic Policy Gradients (TD3). This algorithm sets a new benchmark for performance in continuous robotic control tasks, and we will demonstrate world class performance in the Bipedal Walker environment from the Open AI gym. TD3 is based on the DDPG algorithm, but addresses a number of approximation issues that result in poor performance in DDPG and other actor critic algorithms.

Finally, we will implement the soft actor critic algorithm (SAC). SAC approaches deep reinforcement learning from a totally different angle: by considering entropy maximization, rather than score maximization, as a viable objective. This results in increased exploration by our agent, and world class performance in a number of important Open AI Gym environments.

By the end of the course, you will know the answers to the following fundamental questions in Actor-Critic methods:

  • Why should we bother with actor critic methods when deep Q learning is so successful?

  • Can the advances in deep Q learning be used in other fields of reinforcement learning?

  • How can we solve the explore-exploit dilemma with a deterministic policy?

  • How do we get and deal with overestimation bias in actor-critic methods?

  • How do we deal with the inherent approximation errors in deep neural networks?

This course is for the highly motivated and advanced student. To succeed, you must have prior course work in all the following topics:

  • College level calculus

  • Reinforcement learning

  • Deep learning

The pace of the course is brisk and the topics are at the cutting edge of deep reinforcement learning research, but the payoff is that you will come out knowing how to read research papers and turn them into functional code as quickly as possible. You’ll never have to rely on dodgy medium blog posts again.

Introduction

1
What You Will Learn in this Course
2
Required Background, Software, and Hardware
3
How to Succeed in this Course

Fundamentals of Reinforcement Learning

1
Review of Fundamental Concepts
2
Calculating State Transition Probabilities

This is a simple two state system with two actions. You have to use this information to calculate the state transition probabilities. Please see the handout from lecture 4 for related material.

3
Teaching an AI about Black Jack with Monte Carlo Prediction
4
Teaching an AI How to Play Black Jack with Monte Carlo Control
5
Review of Temporal Difference Learning Methods
6
Teaching an AI about Balance with TD(0) Prediction
7
Teaching an AI to Balance the Cart Pole with Q Learning

Landing on the Moon with Policy Gradients & Actor Critic Methods

1
What's so Great About Policy Gradient Methods?
2
Combining Neural Networks with Monte Carlo: REINFORCE Policy Gradient Algorithm
3
Introducing the Lunar Lander Environment
4
Coding the Agent's Brain: The Policy Gradient Network
5
Coding the Policy Gradient Agent's Basic Functionality
6
Coding the Agent's Learn Function
7
Coding the Policy Gradient Main Loop and Watching our Agent Land on the Moon
8
Actor Critic Learning: Combining Policy Gradients & Temporal Difference Learning
9
Coding the Actor Critic Networks
10
Coding the Actor Critic Agent
11
Coding the Actor Critic Main Loop and Watching Our Agent Land on the Moon

Deep Deterministic Policy Gradients (DDPG): Actor Critic with Continuous Actions

1
Getting up to Speed With Deep Q Learning
2
How to Read and Understand Cutting Edge Research Papers
3
Analyzing the DDPG Paper Abstract and Introduction
4
Analyzing the Background Material
5
What Algorithm Are We Going to Implement?
6
What Results Should We Expect?
7
What Other Solutions are Out There?
8
What Model Architecture and Hyperparameters Do We Need?
9
Handling the Explore-Exploit Dilemma: Coding the OU Action Noise Class
10
Giving our Agent a Memory: Coding the Replay Memory Buffer Class
11
Deep Q Learning for Actor Critic Methods: Coding the Critic Network Class
12
Coding the Actor Network Class
13
Giving our DDPG Agent Simple Autonomy: Coding the Basic Functions of Our Agent
14
Giving our DDPG Agent a Brain: Coding the Agent's Learn Function
15
Coding the Network Parameter Update Functionality
16
Coding the Main Loop and Watching Our DDPG Agent Land on the Moon

Twin Delayed Deep Deterministic Policy Gradients (TD3)

1
Some Tips on Reading this Paper
2
Analyzing the TD3 Paper Abstract and Introduction
3
What Other Solutions Have People Tried?
4
Reviewing the Fundamental Concepts
5
Is Overestimation Bias Even a Problem in Actor-Critic Methods?
6
Why is Variance a Problem for Actor-Critic Methods?
7
What Results Can We Expect?
8
Coding the Brains of the TD3 Agent - The Actor and Critic Network Classes
9
Giving our TD3 Agent Simple Autonomy - Coding the Basic Agent Functionality
10
Giving our TD3 Agent a Brain - Coding the Learn Function
11
Coding the Network Parameter Update Functionality
12
Coding the Main Loop And Watching our Agent Learn to Walk

Soft Actor Critic

1
A Quick Word on the Paper
2
Getting Acquainted With a New Framework
3
Checking Out What Has Been Done Before
4
Inspecting the Foundation of this New Framework
5
Digging Into the Mathematics of Soft Actor Critic
6
Seeing How the New Algorithm Measures Up
7
Coding the Neural Networks
8
Coding the Soft Actor Critic Basic Functionality
9
Coding the Soft Actor Critic Algorithm
10
Coding the Main Loop and Evaluating Our Agent
You can view and review the lecture materials indefinitely, like an on-demand channel.
Definitely! If you have an internet connection, courses on Udemy are available on any device at any time. If you don't have an internet connection, some instructors also let their students download course lectures. That's up to the instructor though, so make sure you get on their good side!
4.4
4.4 out of 5
193 Ratings

Detailed Rating

Stars 5
101
Stars 4
64
Stars 3
16
Stars 2
7
Stars 1
4
30-Day Money-Back Guarantee

Includes

8 hours on-demand video
Full lifetime access
Access on mobile and TV
Certificate of Completion
Modern Reinforcement Learning: Actor-Critic Algorithms
Price:
$218.98 $169

Community

For Professionals

For Businesses

We support Sales, Marketing, Account Management and CX professionals. Learn new skills. Share your expertise. Connect with experts. Get inspired.

Community

Partnership Opportunities

Layer 1
samcx.com
Logo
Register New Account
Compare items
  • Total (0)
Compare
0