Please use this identifier to cite or link to this item: https://www.um.edu.mt/library/oar/handle/123456789/82025
Title: A continuous information gain measure to find the most discriminatory problems for AI benchmarking
Authors: Stephenson, Matthew
Anderson, Damien
Khalifa, Ahmed
Levine, John
Renz, Jochen
Togelius, Julian
Salge, Christoph
Keywords: Computer games -- Design
Multiagent systems
Artificial intelligence
Issue Date: 2020
Publisher: Institute of Electrical and Electronics Engineers
Citation: Stephenson, M., Anderson, D., Khalifa, A., Levine, J., Renz, J., Togelius, J., & Salge, C. (2020). A continuous information gain measure to find the most discriminatory problems for AI benchmarking. 2020 IEEE Congress on Evolutionary Computation (CEC), Glasgow.
Abstract: This paper introduces an information-theoretic method for selecting a subset of problems which gives the most information about a group of problem-solving algorithms. This method was tested on the games in the General Video Game AI (GVGAI) framework, allowing us to identify a smaller set of games that still gives a large amount of information about the abilities of different game-playing agents. This approach can be used to make agent testing more efficient. We can achieve almost as good discriminatory accuracy when testing on only a handful of games as when testing on more than a hundred games, something which is often computationally infeasible. Furthermore, this method can be extended to study the dimensions of the effective variance in game design between these games, allowing us to identify which games differentiate between agents in the most complementary ways.
URI: https://www.um.edu.mt/library/oar/handle/123456789/82025
Appears in Collections:Scholarly Works - InsDG

Files in This Item:
File Description SizeFormat 
A_Continuous_Information_Gain_Measure_to_Find_the_Most_Discriminatory_Problems_for_AI_Benchmarking_2020.pdf
  Restricted Access
925.07 kBAdobe PDFView/Open Request a copy


Items in OAR@UM are protected by copyright, with all rights reserved, unless otherwise indicated.