An evaluation of fast multi-layer perceptron training techniques for games

David G. Robertson

Research output: Chapter in Book/Report/Conference proceedingConference contribution

116 Downloads (Pure)

Abstract

Despite the rise of Artificial Intelligence (AI) in games leading to the adoption of many academic techniques, multi-layer perceptron (MLP) neural networks have bucked this trend and have rarely been used in a game scenario. This is normally due to long training and development times using the standard error back propagation (EBP) training technique. The purpose of this investigation was to compare alternative training techniques to EBP in order to see if they can be used to promote the use of MLP in games.
The application created to serve this purpose was a 2D top down racing game with three different training techniques to control the AI, including EBP, resilient propagation (RPROP) and Random-Minimum Bit Distance Gram Schmidt (RMGS), in which, each training technique was put through three tests.
Through these tests, it was shown that alternative training techniques, although not as accurate as EBP, reduce the training time drastically. The tests also concluded that in a racing game scenario the alternative techniques could also compete with EBP, with the RMGS training technique being the best in every test except accuracy.
This project has shown that MLP could easily be utilised in game scenarios using these alternative methods and would not require the lengthy training times of EBP.
Original languageEnglish
Title of host publicationGAME-ON’2017
Subtitle of host publicationthe 18th International Conference on Intelligent Games and Simulation
EditorsJoseph Kehoe
PublisherEUROSIS
Pages62-66
Number of pages5
ISBN (Print)9789077381991
Publication statusPublished - 24 Aug 2017
EventGAME-ON'2017,18th annual Conference on Simulation and AI in Computer Games - Institute of Technology Carlow, Carlow, Ireland
Duration: 6 Sep 20178 Sep 2017
Conference number: 18
https://www.eurosis.org/cms/?q=node/3661

Conference

ConferenceGAME-ON'2017,18th annual Conference on Simulation and AI in Computer Games
Abbreviated titleGAME-ON'2017
CountryIreland
CityCarlow
Period6/09/178/09/17
Internet address

Fingerprint

Multilayer neural networks
Backpropagation
Artificial intelligence
Neural networks

Cite this

Robertson, D. G. (2017). An evaluation of fast multi-layer perceptron training techniques for games. In J. Kehoe (Ed.), GAME-ON’2017: the 18th International Conference on Intelligent Games and Simulation (pp. 62-66). EUROSIS.
Robertson, David G. / An evaluation of fast multi-layer perceptron training techniques for games. GAME-ON’2017: the 18th International Conference on Intelligent Games and Simulation. editor / Joseph Kehoe. EUROSIS, 2017. pp. 62-66
@inproceedings{aca8540166ec43caa13b9c0f094c3b46,
title = "An evaluation of fast multi-layer perceptron training techniques for games",
abstract = "Despite the rise of Artificial Intelligence (AI) in games leading to the adoption of many academic techniques, multi-layer perceptron (MLP) neural networks have bucked this trend and have rarely been used in a game scenario. This is normally due to long training and development times using the standard error back propagation (EBP) training technique. The purpose of this investigation was to compare alternative training techniques to EBP in order to see if they can be used to promote the use of MLP in games.The application created to serve this purpose was a 2D top down racing game with three different training techniques to control the AI, including EBP, resilient propagation (RPROP) and Random-Minimum Bit Distance Gram Schmidt (RMGS), in which, each training technique was put through three tests.Through these tests, it was shown that alternative training techniques, although not as accurate as EBP, reduce the training time drastically. The tests also concluded that in a racing game scenario the alternative techniques could also compete with EBP, with the RMGS training technique being the best in every test except accuracy.This project has shown that MLP could easily be utilised in game scenarios using these alternative methods and would not require the lengthy training times of EBP.",
author = "Robertson, {David G.}",
year = "2017",
month = "8",
day = "24",
language = "English",
isbn = "9789077381991",
pages = "62--66",
editor = "Joseph Kehoe",
booktitle = "GAME-ON’2017",
publisher = "EUROSIS",

}

Robertson, DG 2017, An evaluation of fast multi-layer perceptron training techniques for games. in J Kehoe (ed.), GAME-ON’2017: the 18th International Conference on Intelligent Games and Simulation. EUROSIS, pp. 62-66, GAME-ON'2017,18th annual Conference on Simulation and AI in Computer Games, Carlow, Ireland, 6/09/17.

An evaluation of fast multi-layer perceptron training techniques for games. / Robertson, David G.

GAME-ON’2017: the 18th International Conference on Intelligent Games and Simulation. ed. / Joseph Kehoe. EUROSIS, 2017. p. 62-66.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - An evaluation of fast multi-layer perceptron training techniques for games

AU - Robertson, David G.

PY - 2017/8/24

Y1 - 2017/8/24

N2 - Despite the rise of Artificial Intelligence (AI) in games leading to the adoption of many academic techniques, multi-layer perceptron (MLP) neural networks have bucked this trend and have rarely been used in a game scenario. This is normally due to long training and development times using the standard error back propagation (EBP) training technique. The purpose of this investigation was to compare alternative training techniques to EBP in order to see if they can be used to promote the use of MLP in games.The application created to serve this purpose was a 2D top down racing game with three different training techniques to control the AI, including EBP, resilient propagation (RPROP) and Random-Minimum Bit Distance Gram Schmidt (RMGS), in which, each training technique was put through three tests.Through these tests, it was shown that alternative training techniques, although not as accurate as EBP, reduce the training time drastically. The tests also concluded that in a racing game scenario the alternative techniques could also compete with EBP, with the RMGS training technique being the best in every test except accuracy.This project has shown that MLP could easily be utilised in game scenarios using these alternative methods and would not require the lengthy training times of EBP.

AB - Despite the rise of Artificial Intelligence (AI) in games leading to the adoption of many academic techniques, multi-layer perceptron (MLP) neural networks have bucked this trend and have rarely been used in a game scenario. This is normally due to long training and development times using the standard error back propagation (EBP) training technique. The purpose of this investigation was to compare alternative training techniques to EBP in order to see if they can be used to promote the use of MLP in games.The application created to serve this purpose was a 2D top down racing game with three different training techniques to control the AI, including EBP, resilient propagation (RPROP) and Random-Minimum Bit Distance Gram Schmidt (RMGS), in which, each training technique was put through three tests.Through these tests, it was shown that alternative training techniques, although not as accurate as EBP, reduce the training time drastically. The tests also concluded that in a racing game scenario the alternative techniques could also compete with EBP, with the RMGS training technique being the best in every test except accuracy.This project has shown that MLP could easily be utilised in game scenarios using these alternative methods and would not require the lengthy training times of EBP.

M3 - Conference contribution

SN - 9789077381991

SP - 62

EP - 66

BT - GAME-ON’2017

A2 - Kehoe, Joseph

PB - EUROSIS

ER -

Robertson DG. An evaluation of fast multi-layer perceptron training techniques for games. In Kehoe J, editor, GAME-ON’2017: the 18th International Conference on Intelligent Games and Simulation. EUROSIS. 2017. p. 62-66