ADS Drinks & Data: Deep Dive into Reinforcement Learning

ADS Drinks & Data: Deep Dive into Reinforcement Learning
Date: Tuesday 25 September 2018
Time: 16:00-18:00
Location: Room C1.110 (1st floor), Science Park 904, UvA
Registration

The ADS Deep Dive sessions are an opportunity to delve into a data science topic in more depth, specifically on Tuesday 25 September we will highlight cutting-edge research on Reinforcement Learning. We will have two speakers, each presenting for 25 minutes plus five minutes discussion time.

Programme
16:00-16:15 – Coffee & walk-in

Introduction & Chair
Zeynep Akata
Assistant Professor (UvA) and Scientific Manager of UvA-Bosch Delta Lab

16:15-16:45
Herke van Hoof
Assistant Professor, Amsterdam Machine Learning Lab (UvA)

“Learning modular strategies with hierarchical reinforcement learning”

Recent years have seen impressive results in learning optimal behaviour in complex systems, for example in the domain of games (Go and Atari) and simulated robots (Mujoco & OpenAI Gym). Many of these results were reached by deep reinforcement learning methods, where deep neural networks are used to model objects of interests. The flexibility of these networks comes at the cost of needing large amount of training data.

However, in many systems, gathering many samples is impracticable. This includes many real-world robotics tasks, as well as systems with humans in the loop. One possible solution to this problem takes the form of learning modular policies, where individual modules can be transferred to related tasks. I will discuss our new approach for hierarchical reinforcement learning that can efficiently learn complex modular policies from experience.

16:45-17:15
Sander Bohte
Researcher Neural Computation (CWI) and Professor of Computational Neuroscience (UvA)

“Efficient Continuous-time Reinforcement Learning”

17:15-18:00: Drinks, snacks & networking

***
The event will be in English.
It is free to attend but please register in advance.
***
Amsterdam Data Science (ADS) accelerates data science and artificial intelligence research by connecting, sharing and showcasing world-class technology, expertise and talent from Amsterdam on a regional, national and international level. Our research enables business and society to better gather, store, analyse and present data in order to gain valuable insights and make informed decisions.
***
Find out more about ADS at https://amsterdamdatascience.nl