September 25, 2017

Reward Functions: Writing for Reinforcement Learning (Video)

Reward Functions: Writing for Reinforcement Learning (Video)

We’ve put together a series of Training Videos to teach customers about reinforcement learning, reward functions, and The Bonsai Platform. Check out Video 1 to get started with an introduction to types of machine learning.

In August we started the release of our Bonsai training videos, a series of five videos to help new customers quickly get up to speed on the Bonsai platform, the Inkling programming language, and reinforcement learning. If you don’t yet have access to the Platform but are looking to learn more, this series will allow you to learn what you’ll need to know before you get started using the platform.

This fourth video is Writing Great Reward Functions where Ross Story, Data Scientist at Bonsai, explains the process of writing reward functions in reinforcement learning. The video assumes that you already have a general understanding of reinforcement learning from the first video in this series, Introduction to Types of Machine Learning.

In this video you will learn the basic topics of reward functions such as shaping, terminal conditions, and negative and positive rewards. Ross then gives a basic and complex example of a reward function, and then concludes with talking about some challenges and advanced topics like hidden state, sequential or conditional tasks, and pitfalls and how to avoid them. Check out the full video below.

 

For more information about how Bonsai uses reinforcement learning, you can watch the next video in the series or read our blog on Mark Hammond’s Deep Reinforcement Learning presentation at GTC this year.

Subscribe to our newsletter to stay up-to-date on our latest product news & more!

SUBSCRIBE NOW