ITU

DLNE: A hybridization of deep learning and neuroevolution for visual control

Research output: Conference Article in Proceeding or Book/Report chapterArticle in proceedingsResearchpeer-review

Standard

DLNE: A hybridization of deep learning and neuroevolution for visual control. / Poulsen, Andreas Precht; Thorhauge, Mark; Funch, Mikkel Hvilshj; Risi, Sebastian.

Computational Intelligence and Games (CIG), 2017 IEEE Conference on. IEEE Press, 2017. p. 256-263.

Research output: Conference Article in Proceeding or Book/Report chapterArticle in proceedingsResearchpeer-review

Harvard

Poulsen, AP, Thorhauge, M, Funch, MH & Risi, S 2017, DLNE: A hybridization of deep learning and neuroevolution for visual control. in Computational Intelligence and Games (CIG), 2017 IEEE Conference on. IEEE Press, pp. 256-263. https://doi.org/10.1109/CIG.2017.8080444

APA

Poulsen, A. P., Thorhauge, M., Funch, M. H., & Risi, S. (2017). DLNE: A hybridization of deep learning and neuroevolution for visual control. In Computational Intelligence and Games (CIG), 2017 IEEE Conference on (pp. 256-263). IEEE Press. https://doi.org/10.1109/CIG.2017.8080444

Vancouver

Poulsen AP, Thorhauge M, Funch MH, Risi S. DLNE: A hybridization of deep learning and neuroevolution for visual control. In Computational Intelligence and Games (CIG), 2017 IEEE Conference on. IEEE Press. 2017. p. 256-263 https://doi.org/10.1109/CIG.2017.8080444

Author

Poulsen, Andreas Precht ; Thorhauge, Mark ; Funch, Mikkel Hvilshj ; Risi, Sebastian. / DLNE: A hybridization of deep learning and neuroevolution for visual control. Computational Intelligence and Games (CIG), 2017 IEEE Conference on. IEEE Press, 2017. pp. 256-263

Bibtex

@inproceedings{3fbb53669e644056890ecaef3213113d,
title = "DLNE: A hybridization of deep learning and neuroevolution for visual control",
abstract = "This paper investigates the potential of combining deep learning and neuroevolution to create a bot for a simple first person shooter (FPS) game capable of aiming and shooting based on high-dimensional raw pixel input. The deep learning component is responsible for visual recognition and translating raw pixels to compact feature representations, while the evolving network takes those features as inputs to infer actions. Two types of feature representations are evaluated in terms of (1) how precise they allow the deep network to recognize the position of the enemy, (2) their effect on evolution, and (3) how well they allow the deep network and evolved network to interface with each other. Overall, the results suggest that combining deep learning and neuroevolution in a hybrid approach is a promising research direction that could make complex visual domains directly accessible to networks trained through evolution.",
author = "Poulsen, {Andreas Precht} and Mark Thorhauge and Funch, {Mikkel Hvilshj} and Sebastian Risi",
year = "2017",
doi = "10.1109/CIG.2017.8080444",
language = "English",
pages = "256--263",
booktitle = "Computational Intelligence and Games (CIG), 2017 IEEE Conference on",
publisher = "IEEE Press",

}

RIS

TY - GEN

T1 - DLNE: A hybridization of deep learning and neuroevolution for visual control

AU - Poulsen, Andreas Precht

AU - Thorhauge, Mark

AU - Funch, Mikkel Hvilshj

AU - Risi, Sebastian

PY - 2017

Y1 - 2017

N2 - This paper investigates the potential of combining deep learning and neuroevolution to create a bot for a simple first person shooter (FPS) game capable of aiming and shooting based on high-dimensional raw pixel input. The deep learning component is responsible for visual recognition and translating raw pixels to compact feature representations, while the evolving network takes those features as inputs to infer actions. Two types of feature representations are evaluated in terms of (1) how precise they allow the deep network to recognize the position of the enemy, (2) their effect on evolution, and (3) how well they allow the deep network and evolved network to interface with each other. Overall, the results suggest that combining deep learning and neuroevolution in a hybrid approach is a promising research direction that could make complex visual domains directly accessible to networks trained through evolution.

AB - This paper investigates the potential of combining deep learning and neuroevolution to create a bot for a simple first person shooter (FPS) game capable of aiming and shooting based on high-dimensional raw pixel input. The deep learning component is responsible for visual recognition and translating raw pixels to compact feature representations, while the evolving network takes those features as inputs to infer actions. Two types of feature representations are evaluated in terms of (1) how precise they allow the deep network to recognize the position of the enemy, (2) their effect on evolution, and (3) how well they allow the deep network and evolved network to interface with each other. Overall, the results suggest that combining deep learning and neuroevolution in a hybrid approach is a promising research direction that could make complex visual domains directly accessible to networks trained through evolution.

U2 - 10.1109/CIG.2017.8080444

DO - 10.1109/CIG.2017.8080444

M3 - Article in proceedings

SP - 256

EP - 263

BT - Computational Intelligence and Games (CIG), 2017 IEEE Conference on

PB - IEEE Press

ER -

ID: 82406769