Workshop / Overview
Humans have been able to achieve incredible feats, such as landing on the moon, exploring the depth of the seas, more than double their life expectancy, harness the power of wind, etc... The main drivers of such advances have usually been the questions "Why?" and "What if?". Curiosity has allowed humans to learn more about their surroundings, sometimes doing things that are counter productive on the short run or for the individual but greatly beneficial on the long run.
The field of Reinforcement Learning explores how a reward can teach an artificial agent to behave a certain way. During this workshop we are going to build our own agent that will be able to learn to play different games, but with a twist: we are going to make it curious! Our agent will not only try to find the short term benefits but will learn new strategies that can give bigger delayed rewards.
Workshop / Outcome
- Build a AI agent able to play various games such as Super Mario
- Learn about Reinforcement Learning
- Learn about Curiosity and Intrinsic Motivation
Workshop / Difficulty
Workshop / Prerequisites
- come with a laptop
- know a bit of python (not mandatory)
- to not loose too much time during the workshop, please follow the 'Build a Docker image from this repository' steps on https://github.com/thibaultcalvayrac/AMLD_artificial_curiosity
Track / Co-organizers
TensorFlow Basics 2019 – Saturday
With Bartek Wołowiec, Megan Ruthven, Ruslan Habalov & Andreas Steiner09:00-16:30 January 26