Reinforcement Learning Workshop

Share on facebook
Share on twitter
Share on linkedin

ToThePoint is co-hosting a reinforcement learning workshop with the folks from The Campus, the 17th of december. This is your chance to grasp an understanding of what Reinforcement Learning is, how to get started with RL and what value you could create for yourself and your (future) company. We’ll be using a load of real-world examples to help you get started.

What can you do with reinforcement learning?

Reinforcement learning, when compared to other kinds of machine learning, gathers its data through trial & error according to a (clever) rewarding policy. In such it opens doors to Machine Learning in many ways (un)supervised learning cannot provide.

What can you do with RL?

How to create a clever rewarding policy.

In our RL workshop the 17th of december, we’ll share all the ins and outs of creating a rewarding policy yourself. For now we’ll leave you with a vague hint of “Value functions” and “the Bellman equations“. See (https://www.youtube.com/watch?v=rx06GkLdzwA) how it works in our physical puzzle solver.

If you take the boat racing example below, you’ll see how important it is to formulate a rewarding policy that helps your machine learn in a way that was intended.

The "agent" has learned to steer in a donut-shaped route, rather than finishing the racetrack. In this way, it is always right on time to gain bonus points. Clever... clever .... clever ...
The “agent” has learned to steer in a donut-shaped route, rather than finishing the racetrack. In this way, it is always right on time to gain bonus points. Clever… clever …. clever …

Testing the policy using a digital twin.

To test your reinforcement learning agent in a real-world scenario, you’ll need a lot of investments. You would not test out a self-driving vehicle in the real world by letting it bump into almost anything in its path without first testing this in a digital twin scenario.

If you have a Virtual Reality version of your office space, you could test out an automatic trash-clearing-robot in this digital world before you test it out in real life.

In the workshop of 17th december, you’ll learn how to create a digital twin for yourself.

MIT makes a pretty good illustration of reinforcement learning showing how their robot becomes a master Rubik’s Cube solver.

Interested in learning more about Reinforcement Learning?

Find out everything there is to know about our workshop in the following PDF or register straight away on the website of The Campus: https://www.thecampus.be/a-conceptual-introduction-to-reinforcement-learning


Also published on Medium.

ToThePoint

Leave a Reply