Machine Learning: Beginner Reinforcement Learning in Python

Name: Machine Learning: Beginner Reinforcement Learning in Python
Rating: 4.3 (164 reviews)

How to teach a neural network to play a game using delayed gratification in 146 lines of Python code

Created byMilo Spencer-Harper

Last updated 1/2020

English

What you'll learn

Machine Learning
Artificial Intelligence
Neural Networks
Reinforcement Learning
Deep Q Learning
OpenAI Gym
Keras
Tensorflow
Bellman Equation

Course content

5 sections • 24 lectures • 1h 44m total length

Introduction to Machine Learning5:20
In this section, I introduce you to machine learning. We define the term and we learn about the three different types of machine learning: supervised, unsupervised and reinforcement learning.
Practise Activity: Machine Learning Quiz
Introduction to Reinforcement Learning7:00
In this section, I introduce you to reinforcement learning, including a history of the field. You will learn the meaning of 5 key terms: agent, environment, state, actions and reward.
Practise Activity: Reinforcement Learning Quiz
The Game: Nchain2:37
In this section, I introduce you to the NChain game. Our challenge is to create an agent intelligent enough to solve this game.
Practise Activity: Nchain Quiz
Reviewing your Practise Activity0:41

Windows: Google Colab2:30
For Windows users, I show you how to use Google Colab to run your code in the cloud.
Mac/Linux: Installing Python, Sublime and the Terminal4:33
In this section, I show Mac and Linux users how to install Python 3 and Sublime, as well as how to use the Terminal. If you are using Windows, use Google Colab instead.
Mac/Linux: Installing the Open AI gym3:28
In this section, I show Mac and Linux users how to instal the Open AI gym library. The library is not compatible with Windows, so if you use Windows use Google Colab instead.
Coding: Random Exploration7:12
We get familiar with the NChain game, by creating an agent which randomly explores..
Coding: Refactoring to a Class2:29
We refactor our code into a class, ready for the next section.
Coding: Reward Table6:44
We give the agent a memory using a reward table, which maps the cumulative reward for each state-action combination.
Coding: Choose the Highest Reward6:26
We create an agent, which chooses the highest reward. We find it gets locked into a sub-optimal strategy. It is not intelligent enough to accept delayed gratification to gain the long-term reward. Building this low intelligence agent helps us understand the challenge we have set ourselves, and leads into a better approach, using Q Learning. By contrasting the two approaches we can understand why Q Learning is so revolutionary.

Requirements

Basic knowledge of Python

Description

This course is designed for beginners to machine learning. Some of the most exciting advances in artificial intelligence have occurred by challenging neural networks to play games. I will introduce the concept of reinforcement learning, by teaching you to code a neural network in Python capable of delayed gratification.

We will use the NChain game provided by the Open AI institute. The computer gets a small reward if it goes backwards, but if it learns to make short term sacrifices by persistently pressing forwards it can earn a much larger reward. Using this example I will teach you Deep Q Learning - a revolutionary technique invented by Google DeepMind to teach neural networks to play chess, Go and Atari.

Who this course is for:

Anyone interested in machine learning

Machine Learning: Beginner Reinforcement Learning in Python

What you'll learn

Explore related topics

Course content

Introduction4 lectures • 16min

Creating your Agent and Environment7 lectures • 33min

Q Learning5 lectures • 25min

Neural Networks3 lectures • 8min

Deep Q Learning5 lectures • 23min

Requirements

Description

Who this course is for: