arXiv preprint

Early Class Exclusion in Hyperdimensional Computing

Rémy Duboucheix, Mohsen Asghari, Sébastien Le Beux, Otmane Ait Mohamed, Ron Mankarious

27th International Symposium on Quality Electronic Design (ISQED) · 2026 · arXiv:https://ieeexplore.ieee.org/document/11534675

Accélérateurs d'IA embarquéeCalcul approximatif

27th International Symposium on Quality Electronic Design (ISQED) 2026 Rémy Duboucheix, Mohsen Asghari, Sébastien Le Beux, Otmane Ait Mohamed, Ron Mankarious

Résumé

Hyperdimensional computing (HDC) is a lightweight machine-learning paradigm in which data are encoded as high-dimensional hypervectors, enabling error-resilient and hardware-friendly computation. This makes HDC well suited for low-resource embedded systems that must balance performance and energy. However, existing HDC accelerators require full similarity checks between query and class hypervectors, incurring significant computational and energy cost. We propose an HDC algorithm that progressively excludes low-similarity classes by iterating over segmented hypervectors, reducing the number of required comparisons without storing full class representations. Our RTL accelerator reduces execution time by 3.7× and energy consumption by 28.7%, with only a 3.2% accuracy loss.

Citation

Si vous citez ces travaux, merci d'utiliser l'entrée ci-dessous. Vous pouvez copier le BibTeX dans le presse-papier via le bouton en haut de page.

@misc{rmy2026httpsieeexploreieeeorgdocument11534675,
  title         = {Early Class Exclusion in Hyperdimensional Computing},
  author        = {Rémy Duboucheix and Mohsen Asghari and Sébastien Le Beux and Otmane Ait Mohamed and Ron Mankarious},
  year          = {2026},
  eprint        = {https://ieeexplore.ieee.org/document/11534675},
  archivePrefix = {arXiv},
  primaryClass  = {cs.AR}
}

Remerciements

Ces travaux ont été soutenus en partie par le Conseil de recherches en sciences naturelles et en génie du Canada (CRSNG) et par le Fonds de recherche du Québec — Nature et technologies (FRQNT).