Data Science and Machine Learning in Games

Video games, from computers to mobile phones, are a massive phenomenon with hundreds of millions of people playing every day. Nowadays, most of games on every platform collect data about the player’s behaviour, what are the challenges involved? How can data be used to make better games and more successful game companies?

Dr. Sebastian Risi from the IT University of Copenhagen and Johan Kjaergaard from Tactile Games will talk about their experiences and discuss the future of data science and machine learning in games.

Agenda

• Intro to data science and machine learning in games (~10 min)
• Data Engineering in Games - A practical approach (~45 min)
• Break (~15 min)
• Tactile thesis prize (~15 min)
• Why Video Games could be the Key towards Artificial Intelligence (~45 min)

Talks

• Data Engineering in Games - A practical approach
We want to make business decisions based on data. Significant overhead and daily maintenance should not be necessary to aggregate on any data. Different solutions come with strengths and weaknesses so we decided on a custom solution made of decentralized parts with the data ending up in a fully maintained external data warehouse.
Key objectives was scalability, easy maintenance and ease of access for all users in our organisation on even raw telemetry data. The data warehouse should be able to scale efficiently across multiple games with millions of users and billions of rows.
This talk will cover parts of our setup to collect telemetry data sent from clients, how that stays dynamic and customizable at all times, how to aggregate on the data in steps based on a containerized application with minimal overhead and maintenance in a decentralized fashion, and in the end how we make useful decisions based on the aggregated data.

• Why Video Games could be the Key towards Artificial Intelligence
- Many recent breakthroughs in Artificial Intelligence (AI) would potentially not have been possible without the help of games. Especially video games with their complex worlds and large action spaces offer unique testbeds for advanced AI methods.
Breakthroughs that have first been invented to beat the best human players in games such as Go or Poker are now being applied in many different domains such as healthcare or improving the efficiency of power grids.
While machines have surpassed human-level performance in many games already, some video games such as StarCraft are still too complex for even the most advanced AI methods. In this lecture I will tell you about the recent progress in this field and how letting machines play games could ultimately allow them to learn and to think more like humans.

Venue

The event will be hosted by the IT University of Copenhagen, food and drinks will be sponsored by Tactile Games.

Time and location

Wednesday, January 30th 2019
From 6:00 PM to 8:30 PM
IT University of Copenhagen
Rued Langgaards Vej 7
2300 Copenhagen

We will be in auditorium 2!