VUB AI Webinar Series: Reinforcement Learning

Reinforcement Learning: from the lab to the real world

In the framework of the Flanders AI Research Program, VUB organizes a webinar series on AI. The first webinar (a tutorial-style presentation organized digitally through Zoom) will focus on reinforcement learning, one of the core expertises of the VUB AI Lab. It aims to engage with Flemish companies to stimulate the adoption of AI in Flanders.

When? Thursday December 3rd, 15h30-17h
Where? If you register below, you will receive a link to follow the webinar a few days in advance. Registration is free.
Target audience: companies with no or limited knowledge in AI and reinforcement learning who want to learn more about the opportunities for their business


Abstract: Reinforcement Learning (RL) allows an artificial agent to learn how to perform a task. RL learns from experience and is radically different from better-known machine learning techniques typically used in data mining. It borrows ideas from psychology on operant conditioning (not unlike training a dog with cookies) and addresses problems for which the user specifies what has to be achieved, rather how it can be done. While RL received worldwide attention after Google’s AlphaGo beat the world champion Go in 2016, the field has been showcasing interesting applications since the 90s. Today, RL found its way outside the labs and is widely applied to optimize software and hardware control systems: robots, smart energy grids, autonomous cars, manufacturing planning, datacenter cost-of-ownership reduction, optimalisation, etc.

In this webinar, we will explain the basic concepts of RL, present early success stories and recent advancements, including how they can be leveraged by the industry today. While early RL algorithms were evaluated on toy problems, in recent years advanced methods have been developed, which now enable RL to be deployed as part of real-world solutions useful for industry. Finally, we will go past the technical aspects and address challenges such as transparency and explainability, which have gained a lot of importance since the European regulation on GDPR came into effect.

The VUB AI Lab, headed by Prof. Ann Nowé, has been performing research on reinforcement learning for more than 30 years. Today the lab is focusing on various challenges in reinforcement learning, such as increasing learning speed, integration of deep learning models, increasing cooperation between AI systems, dealing with multi-objective criteria, explainability and safety. Within the Flanders AI Research Program, the VUB AI Lab is leading the challenge on multi-agent collaborative AI, investigating coordination mechanisms for AI systems. Reinforcement learning plays an important role, as these methods need to be highly adaptive in order to cope with rapidly changing environments.