EUVIMEDEUVIMED entdecken
Crossref

Optimized Kernels in MobileNet Architectures for Computer Chess

Raphaël Mathiot, Olivier Goudet, Tristan Cazenave

ICGA Journal: The Journal of the Computer Games Community · 2026-05 · 48 (2) · 38-61

DOI: 10.1177/13896911261467835 ↗

Kurz gefasst

Fachartikel in der Zeitschrift ICGA Journal: The Journal of the Computer Games Community, online erschienen am 27.07.2026. Verfasst von Raphaël Mathiot und 2 weiteren Autor:innen. Die Arbeit nennt 15 Literaturangaben. Gefördert von: GENCI.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This work investigates the use of various Residual and MobileNet architectures within the Leela Chess Zero (Lc0) engine. In particular, it examines convolutional layers using specific kernels inspired by chess piece movement patterns. The most effective configuration, combining knight, rook, and bishop filters, is further optimized at a low implementation level to speed up network inference during move search. All network variants are trained on the recent T80 dataset, consisting of self-played games generated by Lc0. Their relative performance is then evaluated through a series of tournaments conducted under different time controls, including bullet, blitz, and rapid formats.

Förderung

  • GENCI

Zitieren

APA

Mathiot, R., Goudet, O., & Cazenave, T. (2026). Optimized Kernels in MobileNet Architectures for Computer Chess. ICGA Journal: The Journal of the Computer Games Community, 48(2), 38-61. https://doi.org/10.1177/13896911261467835

Vancouver

Mathiot R, Goudet O, Cazenave T. Optimized Kernels in MobileNet Architectures for Computer Chess. ICGA Journal: The Journal of the Computer Games Community. 2026;48(2):38-61. doi:10.1177/13896911261467835

BibTeX
@article{Mathiot2026,
  author = {Mathiot, Raphaël and Goudet, Olivier and Cazenave, Tristan},
  title = {Optimized Kernels in MobileNet Architectures for Computer Chess},
  journal = {ICGA Journal: The Journal of the Computer Games Community},
  year = {2026},
  volume = {48},
  number = {2},
  pages = {38--61},
  doi = {10.1177/13896911261467835}
}
Kostenlos starten →