Skip to content

Meeting 27 October 2017

Jump to bottom
SimonCraenen edited this page Oct 27, 2017 · 3 revisions

Meeting Minutes - Project Group 8

Location:  room 2.002 Bouillonstraat
Date:   27 October 2017
Time:   13:00-14:00

Attendance

  • Joshua Scheidt
  • Marciano Geijselaers
  • Timo Raff
  • Simon Craenen
  • Mark Winands
  • Kurt Driessens
  • Jialin Liu

Agenda Questions

  • What are the possible methods we can use?
  • Can we train multiple neural networks depending on the category of a game? Is this still general? We would decide in the first moments of the game what neural network to use.
  • Is it possible to use a limited area view from the avatars perspective in all types of games? (It could be a square around the player or a rectangle around the player and its field of view)

Agenda notes:

Notes with Kurt: deictic field of view.

The 5 learning minutes do not include the response time of the server, it's purely the CPU time of the learning agent. Quickly firing events even randomly can gain a lot of observations. We can play the game as a human ourselves in the planning track, GVGAI src tracks singlePlayer Test.java Classify type of game and choose strategy afterwards.

Clone this wiki locally