GDC is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

View, browse and sort the ever-growing list of sessions by pass type, topic, and format. With this Session Scheduler, you can build your schedule in advance and access it during the show via export or with the GDC Mobile App, once live. Sessions do fill up so please arrive early to sessions that you would like to attend. Adding a session to your schedule does not guarantee you a seat.

Machine Learning Summit: A.N.N.A. Reinforcement Learning In Racing Games

Giuseppe Campana (Lead Machine Learning Programmer, Milestone Srl)

Pass Type: All Access Pass, Core+Summits Pass, Summits Pass - Get your pass now!

Topic: Programming

Format: Session

Vault Recording: TBD

Audience Level: Intermediate

Reinforcement learning is the future of N.P.C. in videogames but squeezing from it even acceptable results it's still a hard task. You set the rules of the game, you set the goal, then an optimizer explores the weights space to find a good (in your hope the best) model. But most time you find out that it just didn't work. Maybe the goal you have chosen is not the one you meant, or maybe that the path to it is not smooth enough. And you can't debug trained agents, nor they tell you the reasons why they are not behaving as you expected. This talk presents the successful case of MotoGP™ 19, giving an insight into how the classical difficulties of machine learning, like local minima and overfitting, materialize in the context of racing games.