We are very excited to have Gema Parreño over in Reading all the way from Madrid to show us how to create a deep reinforcement learning (DRL) agent for StarCraftII!!
She has a lot of experience in this area and is very kindly sharing it by creating a detailed codelab; after an introductory talk she will walk us through the codelab so we can actually get some hands-on experience. For those who want to follow please bring your laptops.
If you don't know anything about RL this is a really good way to find out how it works, why people work on it, and what the challenges are so come along and listen.
StarCraft is a challenging real time strategy game and is used as a testbed for many research groups working on reinforcement learning and various aspects of ML. You may have heard of AlphaStar ( if you want to read more see https://deepmind.com/blog/alphastar-mastering-real-time-strategy-game-starcraft-ii/) recently from Deepmind.
This talk and lab will give you some idea of how to get started with a complex environment where you can try out your own ideas. Gema will also be able to give you insights into what is hard and how people try to tackle these hard problems.
Agenda:
18:30 - Food, drinks, networking
19:00 - Gema (@soygema) - Introduction to reinforcement learning and the StarCraft II Learning environment
19:50 Short break
20:00 - Gema - Codelab - a hands on approach with the Starcraft II environment.
21: 00 - Event close
BIO : Gema Parreño is Data Scientist at BBVA Innovation Labs. She designed DeepAsteroid, a Neural net architecture able to predict asteroids impact on earth . The project was selected among the 25th most innovative in the world and in between the 5th of the best use of data in NASA Space Apps Challenge 2016 and selected as one of the cases of use for TensorFlow in Google I/O 2017. She currently contributes to several open source projects having machine learning at their core.
More at https://github.com/soygema
https://twitter.com/SoyGema
Requirements for codelab:
Starcraft takes quite a while to install so you need to do that before you get here if you want to participate, else you can just watch and make notes and try it at home!
GitHub
GDG Organizer
GDG Organizer
NA
GDG Organizer
GDG Organizer
GDG Organizer