Reinforcement Learning (RL)

15:55 Walk In
16:00 Introduction & Welcome
16:05 Talk #1: Herke van Hoof (UvA, ICAI, AMLab, Bosch Delta Lab): Structure and symmetry in Reinforcement Learning
16:20 Discussion
16:25 Talk #2: Sami Jullien (UvA, ICAI, AIRLab): Modelling and Optimizing in a Partially Observable Environment
16:40 Discussion
16:45 Talk #3: Samir Araújo de Souza (Amazon Web Services): Explore a reference architecture + Open Source frameworks like RLLib
17:00 Discussion
17:05 End

Herke van Hoof is currently assistant professor at the University of Amsterdam in the Netherlands. He is part of the Amlab headed by Professor Max Welling as well as the UvA-Bosch Delta lab. Herke works on machine learning for autonomous robots in perceptually challenging environments. For robots to master a wide array of tasks, machine learning is indispensable, but it is equally important that such tasks can be learned in non-standardized and unstructured environments such as homes or hospitals. Learning tasks in such complicated environment puts additional demands on algorithms for machine learning, perception, and control.

Before joining the University of Amsterdam, Herke van Hoof was a postdoc at McGill University in Montreal, Canada, where he worked with Professors Joelle Pineau, Dave Meger, and Gregory Dudek. He obtained his PhD at TU Darmstadt, Germany, under the supervision of Professor Jan Peters, where he graduated in November 2016. Herke got his bachelor and master degrees in Artificial Intelligence at the University of Groningen in the Netherlands.

Structure and symmetry in Reinforcement Learning:
Reinforcement learning methods have been shown to yield impressive results in e.g. game playing or robot control. However, they are infamously data hungry. This means such algorithms need a long training time, and possibly extensive access to expensive and sensitive hardware such as a robot.
Thus, it is important to exploit knowledge about the structure of tasks. One example is symmetries, which occur in many tasks. For example, a robot trained to pick up objects on its right side with its right arm, could execute the same motion with its left arm to pick up an object on its left. In this talk, I will discuss our recent work, in which we propose a scheme for exploiting such structure in reinforcement learning policies and value networks, and show that it aids convergence speeds in various tasks.

Sami is a PhD student at the AI for Retail Lab and the Information Retrieval Lab at the University of Amsterdam. He loves to apply Artificial Intelligence to the industrial domain, mainly in the fields of supply chain and energy. With a background in statistical engineering, computer science and economics, his current research is on waste reduction in grocery stores via reinforcement learning-based algorithms.
Outside of work, he’s mainly tuning his cooking recipes, language skills, DIY projects and cycling itineraries.

Modelling and Optimizing in a Partially Observable Environment:
In this talk, I will present the way I approach building solutions to solve applied problems. More specifically, I will use the example of my grocery store simulation, were an agent passes order from a warehouse to a grocery store, while trying to optimize sales and reduce generated waste.

Samir Araújo is an AI/ML Solutions Architect at AWS. He helps customers creating AI/ML solutions which solve their business challenges using the AWS platform. He has been working on several AI/ML projects related to computer vision, natural language processing, forecasting, ML at the edge, and more. He likes playing with hardware and automation projects in his free time, and he has a particular interest for robotics.

In this quick session, you’ll see how some AWS Customers are applying RL to solve real business problems. Also, let’s explore a reference architecture to learn more about how they are building complex environments at scale that combine AWS services + Open Source frameworks like RLLib for training agents that operate robots, human assistants and traders in the energy market.