This course is all about the application of **deep learning** and **neural networks** to **reinforcement learning**.

If you’ve taken my first reinforcement learning class, then you know that reinforcement learning is on the bleeding edge of what we can do with AI.

Specifically, the combination of deep learning with reinforcement learning has led to AlphaGo beating a world champion in the strategy game Go, it has led to self-driving cars, and it has led to machines that can play video games at a superhuman level.

Reinforcement learning has been around since the 70s but none of this has been possible until now.

The world is changing at a very fast pace. The state of California is changing their regulations so that self-driving car companies can test their cars without a human in the car to supervise.

We’ve seen that reinforcement learning is an entirely different kind of machine learning than supervised and unsupervised learning.

Supervised and unsupervised**machine learning** algorithms are for analyzing and making predictions about data, whereas reinforcement learning is about training an agent to interact with an environment and maximize its reward.

Unlike supervised and unsupervised learning algorithms, reinforcement learning agents have an impetus - they want to reach a goal.

This is such a fascinating perspective, it can even make supervised / unsupervised machine learning and "**data science**" seem boring in hindsight. Why train a neural network to learn about the data in a database, when you can train a neural network to interact with the real-world?

While deep reinforcement learning and AI has a lot of potential, it also carries with it huge risk.

**Bill Gates** and **Elon Musk** have made public statements about some of the risks that AI poses to economic stability and even our existence.

As we learned in my first reinforcement learning course, one of the main principles of training reinforcement learning agents is that there are unintended consequences when training an AI.

AIs don’t think like humans, and so they come up with novel and non-intuitive solutions to reach their goals, often in ways that surprise domain experts - humans who are the best at what they do.

**OpenAI** is a non-profit founded by Elon Musk, Sam Altman (Y Combinator), and others, in order to ensure that AI progresses in a way that is beneficial, rather than harmful.

Part of the motivation behind OpenAI is the existential risk that AI poses to humans. They believe that open collaboration is one of the keys to mitigating that risk.

One of the great things about OpenAI is that they have a platform called the**OpenAI Gym**, which we’ll be making heavy use of in this course.

It allows anyone, anywhere in the world, to train their reinforcement learning agents in standard environments.

In this course, we’ll build upon what we did in the last course by working with more complex environments, specifically, those provided by the OpenAI Gym:

To train effective learning agents, we’ll need new techniques.

We’ll extend our knowledge of temporal difference learning by looking at the TD Lambda algorithm, we’ll look at a special type of neural network called the RBF network, we’ll look at the policy gradient method, and we’ll end the course by looking at Deep Q-Learning (DQN) and A3C (Asynchronous Advantage Actor-Critic).

Thanks for reading, and I’ll see you in class!

Suggested Prerequisites:

Tips for success:

If you’ve taken my first reinforcement learning class, then you know that reinforcement learning is on the bleeding edge of what we can do with AI.

Specifically, the combination of deep learning with reinforcement learning has led to AlphaGo beating a world champion in the strategy game Go, it has led to self-driving cars, and it has led to machines that can play video games at a superhuman level.

Reinforcement learning has been around since the 70s but none of this has been possible until now.

The world is changing at a very fast pace. The state of California is changing their regulations so that self-driving car companies can test their cars without a human in the car to supervise.

We’ve seen that reinforcement learning is an entirely different kind of machine learning than supervised and unsupervised learning.

Supervised and unsupervised

Unlike supervised and unsupervised learning algorithms, reinforcement learning agents have an impetus - they want to reach a goal.

This is such a fascinating perspective, it can even make supervised / unsupervised machine learning and "

While deep reinforcement learning and AI has a lot of potential, it also carries with it huge risk.

As we learned in my first reinforcement learning course, one of the main principles of training reinforcement learning agents is that there are unintended consequences when training an AI.

AIs don’t think like humans, and so they come up with novel and non-intuitive solutions to reach their goals, often in ways that surprise domain experts - humans who are the best at what they do.

Part of the motivation behind OpenAI is the existential risk that AI poses to humans. They believe that open collaboration is one of the keys to mitigating that risk.

One of the great things about OpenAI is that they have a platform called the

It allows anyone, anywhere in the world, to train their reinforcement learning agents in standard environments.

In this course, we’ll build upon what we did in the last course by working with more complex environments, specifically, those provided by the OpenAI Gym:

- CartPole
- Mountain Car
- Atari games

To train effective learning agents, we’ll need new techniques.

We’ll extend our knowledge of temporal difference learning by looking at the TD Lambda algorithm, we’ll look at a special type of neural network called the RBF network, we’ll look at the policy gradient method, and we’ll end the course by looking at Deep Q-Learning (DQN) and A3C (Asynchronous Advantage Actor-Critic).

Thanks for reading, and I’ll see you in class!

Suggested Prerequisites:

- calculus
- matrix arithmetic (adding, multiplying)
- probability
- Python coding: if/else, loops, lists, dicts, sets
- Numpy coding: matrix and vector operations, loading a CSV file
- linear regression, logistic regression
- neural networks and backpropagation
- Can write a feedforward neural network in Theano and TensorFlow
- Can write a convolutional neural network and recurrent neural network
- Markov Decision Proccesses (MDPs)
- Know how to implement Dynamic Programming, Monte Carlo, and Temporal Difference Learning to solve MDPs

Tips for success:

- Use the video speed changer! Personally, I like to watch at 2x.
- Take handwritten notes. This will drastically increase your ability to retain the information.
- Write down the equations. If you don't, I guarantee it will just look like gibberish.
- Ask lots of questions on the discussion board. The more the better!
- Don't get discouraged if you can't solve every exercise right away. Sometimes it'll take hours, days, or maybe weeks!
- Write code yourself, this is an applied course! Don't be a "couch potato".

- Introduction and Outline (07:23) (FREE preview available)
- Where to get the Code (05:01)
- How to Succeed in this Course (05:52)
- Tensorflow or Theano - Your Choice! (04:10)

- Reinforcement Learning Section Introduction (06:34)
- Elements of a Reinforcement Learning Problem (20:18)
- States, Actions, Rewards, Policies (09:24)
- Markov Decision Processes (MDPs) (10:07)
- The Return (04:56)
- Value Functions and the Bellman Equation (09:53)
- What does it mean to “learn”? (07:18)
- Solving the Bellman Equation with Reinforcement Learning (pt 1) (09:49)
- Solving the Bellman Equation with Reinforcement Learning (pt 2) (12:01)
- Epsilon-Greedy (06:09)
- Q-Learning (14:15)
- How to Learn Reinforcement Learning (05:56)
- Suggestion Box (03:03)

- Review Intro (02:41)
- Review of Markov Decision Processes (07:48)
- Review of Dynamic Programming (04:12)
- Review of Monte Carlo Methods (03:56)
- Review of Temporal Difference Learning (04:42)
- Review of Approximation Methods for Reinforcement Learning (02:20)
- Review of Deep Learning (06:47)
- Suggestion Box (03:03)

- OpenAI Gym Tutorial (05:43)
- Random Search (05:49)
- Saving a Video (02:18)
- CartPole with Bins (Theory) (03:52)
- CartPole with Bins (Code) (06:25)
- RBF Neural Networks (10:26)
- RBF Networks with Mountain Car (Code) (05:28)
- RBF Networks with CartPole (Theory) (01:55)
- RBF Networks with CartPole (Code) (03:12)
- Theano Warmup (03:04)
- Tensorflow Warmup (02:26)
- Plugging in a Neural Network (03:40)
- OpenAI Gym Section Summary (03:28)

- N-Step Methods (03:14)
- N-Step in Code (03:41)
- TD Lambda (07:36)
- TD Lambda in Code (03:00)
- TD Lambda Summary (02:22)

- Policy Gradient Methods (11:39)
- Policy Gradient in TensorFlow for CartPole (07:20)
- Policy Gradient in Theano for CartPole (04:15)
- Continuous Action Spaces (04:17)
- Mountain Car Continuous Specifics (04:12)
- Mountain Car Continuous Theano (07:32)
- Mountain Car Continuous Tensorflow (08:07)
- Mountain Car Continuous Tensorflow (v2) (06:11)
- Mountain Car Continuous Theano (v2) (07:31)
- Policy Gradient Section Summary (01:36)

- Deep Q-Learning Intro (03:52)
- Deep Q-Learning Techniques (09:14)
- Deep Q-Learning in Tensorflow for CartPole (05:10)
- Deep Q-Learning in Theano for CartPole (04:48)
- Additional Implementation Details for Atari (05:36)
- Pseudocode and Replay Memory (06:15)
- Deep Q-Learning in Tensorflow for Breakout (23:47)
- Deep Q-Learning in Theano for Breakout (23:55)
- Partially Observable MDPs (04:52)
- Deep Q-Learning Section Summary (04:45)

- A3C - Theory and Outline (16:30)
- A3C - Code pt 1 (Warmup) (06:28)
- A3C - Code pt 2 (06:28)
- A3C - Code pt 3 (07:36)
- A3C - Code pt 4 (18:02)
- A3C - Section Summary (02:06)
- Course Summary (04:57)

- Introduction and Outline (09:58)
- Where to get the Code (09:21)
- How to Succeed in this Course (08:45)

- Review Intro (02:42)
- Review of Markov Decision Processes (07:48)
- Review of Dynamic Programming (04:13)
- Review of Monte Carlo Methods (03:56)
- Review of Temporal Difference Learning (04:42)
- Review of Approximation Methods for Reinforcement Learning (02:20)
- Review of Deep Learning (06:47)

- Theano Basics: Variables, Functions, Expressions, Optimization (07:47)
- Building a neural network in Theano (09:17)
- TensorFlow Basics: Variables, Functions, Expressions, Optimization (07:27)
- Building a neural network in TensorFlow (09:43)

- Windows-Focused Environment Setup 2018 (20:21)
- How to install Numpy, Scipy, Matplotlib, Pandas, IPython, Theano, and TensorFlow (17:33)

- How to Code Yourself (part 1) (15:55)
- How to Code Yourself (part 2) (09:24)
- Proof that using Jupyter Notebook is the same as not using it (12:29)
- Python 2 vs Python 3 (04:38)
- Is Theano Dead? (10:04)

- How to Succeed in this Course (Long Version) (10:25)
- Is this for Beginners or Experts? Academic or Practical? Fast or slow-paced? (22:05)
- What order should I take your courses in? (part 1) (11:19)
- What order should I take your courses in? (part 2) (16:07)

- What is the Appendix? (02:48)
- Where to get discount coupons and FREE deep learning material (05:31)