The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
Le calcul par réservoir (RC) est une alternative intéressante aux modèles d'apprentissage automatique en raison de son processus de formation peu coûteux en termes de calcul et de sa simplicité. Dans ce travail, nous proposons EnsembleBloomCA, qui utilise des automates cellulaires (CA) et un filtre Bloom d'ensemble pour organiser un système RC. Contrairement à la plupart des systèmes RC existants, EnsembleBloomCA élimine tous les calculs à virgule flottante et les multiplications entières. EnsembleBloomCA adopte CA comme réservoir dans le système RC car il peut être implémenté en utilisant uniquement des opérations binaires et est donc économe en énergie. La riche dynamique de modèles créée par CA peut mapper l'entrée d'origine dans un espace de grande dimension et fournir plus de fonctionnalités au classificateur. En utilisant un filtre Bloom d'ensemble comme classificateur, les caractéristiques fournies par le réservoir peuvent être efficacement mémorisées. Notre expérience a révélé que l'application du mécanisme d'ensemble au filtre Bloom entraînait une réduction significative du coût de la mémoire pendant la phase d'inférence. En comparaison avec Bloom WiSARD, un des ouvrages de référence les plus pointus, le EnsembleBloomCA Le modèle permet une réduction de 43 fois du coût de la mémoire tout en conservant la même précision. Notre implémentation matérielle a également démontré que EnsembleBloomCA obtenu des réductions de surface et de puissance respectivement de plus de 23× et 8.5×.
Dehua LIANG
Osaka University
Jun SHIOMI
Osaka University
Noriyuki MIURA
Osaka University
Masanori HASHIMOTO
Kyoto University
Hiromitsu AWANO
Kyoto University
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copier
Dehua LIANG, Jun SHIOMI, Noriyuki MIURA, Masanori HASHIMOTO, Hiromitsu AWANO, "A Hardware Efficient Reservoir Computing System Using Cellular Automata and Ensemble Bloom Filter" in IEICE TRANSACTIONS on Information,
vol. E105-D, no. 7, pp. 1273-1282, July 2022, doi: 10.1587/transinf.2021EDP7203.
Abstract: Reservoir computing (RC) is an attractive alternative to machine learning models owing to its computationally inexpensive training process and simplicity. In this work, we propose EnsembleBloomCA, which utilizes cellular automata (CA) and an ensemble Bloom filter to organize an RC system. In contrast to most existing RC systems, EnsembleBloomCA eliminates all floating-point calculation and integer multiplication. EnsembleBloomCA adopts CA as the reservoir in the RC system because it can be implemented using only binary operations and is thus energy efficient. The rich pattern dynamics created by CA can map the original input into a high-dimensional space and provide more features for the classifier. Utilizing an ensemble Bloom filter as the classifier, the features provided by the reservoir can be effectively memorized. Our experiment revealed that applying the ensemble mechanism to the Bloom filter resulted in a significant reduction in memory cost during the inference phase. In comparison with Bloom WiSARD, one of the state-of-the-art reference work, the EnsembleBloomCA model achieves a 43× reduction in memory cost while maintaining the same accuracy. Our hardware implementation also demonstrated that EnsembleBloomCA achieved over 23× and 8.5× reductions in area and power, respectively.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2021EDP7203/_p
Copier
@ARTICLE{e105-d_7_1273,
author={Dehua LIANG, Jun SHIOMI, Noriyuki MIURA, Masanori HASHIMOTO, Hiromitsu AWANO, },
journal={IEICE TRANSACTIONS on Information},
title={A Hardware Efficient Reservoir Computing System Using Cellular Automata and Ensemble Bloom Filter},
year={2022},
volume={E105-D},
number={7},
pages={1273-1282},
abstract={Reservoir computing (RC) is an attractive alternative to machine learning models owing to its computationally inexpensive training process and simplicity. In this work, we propose EnsembleBloomCA, which utilizes cellular automata (CA) and an ensemble Bloom filter to organize an RC system. In contrast to most existing RC systems, EnsembleBloomCA eliminates all floating-point calculation and integer multiplication. EnsembleBloomCA adopts CA as the reservoir in the RC system because it can be implemented using only binary operations and is thus energy efficient. The rich pattern dynamics created by CA can map the original input into a high-dimensional space and provide more features for the classifier. Utilizing an ensemble Bloom filter as the classifier, the features provided by the reservoir can be effectively memorized. Our experiment revealed that applying the ensemble mechanism to the Bloom filter resulted in a significant reduction in memory cost during the inference phase. In comparison with Bloom WiSARD, one of the state-of-the-art reference work, the EnsembleBloomCA model achieves a 43× reduction in memory cost while maintaining the same accuracy. Our hardware implementation also demonstrated that EnsembleBloomCA achieved over 23× and 8.5× reductions in area and power, respectively.},
keywords={},
doi={10.1587/transinf.2021EDP7203},
ISSN={1745-1361},
month={July},}
Copier
TY - JOUR
TI - A Hardware Efficient Reservoir Computing System Using Cellular Automata and Ensemble Bloom Filter
T2 - IEICE TRANSACTIONS on Information
SP - 1273
EP - 1282
AU - Dehua LIANG
AU - Jun SHIOMI
AU - Noriyuki MIURA
AU - Masanori HASHIMOTO
AU - Hiromitsu AWANO
PY - 2022
DO - 10.1587/transinf.2021EDP7203
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E105-D
IS - 7
JA - IEICE TRANSACTIONS on Information
Y1 - July 2022
AB - Reservoir computing (RC) is an attractive alternative to machine learning models owing to its computationally inexpensive training process and simplicity. In this work, we propose EnsembleBloomCA, which utilizes cellular automata (CA) and an ensemble Bloom filter to organize an RC system. In contrast to most existing RC systems, EnsembleBloomCA eliminates all floating-point calculation and integer multiplication. EnsembleBloomCA adopts CA as the reservoir in the RC system because it can be implemented using only binary operations and is thus energy efficient. The rich pattern dynamics created by CA can map the original input into a high-dimensional space and provide more features for the classifier. Utilizing an ensemble Bloom filter as the classifier, the features provided by the reservoir can be effectively memorized. Our experiment revealed that applying the ensemble mechanism to the Bloom filter resulted in a significant reduction in memory cost during the inference phase. In comparison with Bloom WiSARD, one of the state-of-the-art reference work, the EnsembleBloomCA model achieves a 43× reduction in memory cost while maintaining the same accuracy. Our hardware implementation also demonstrated that EnsembleBloomCA achieved over 23× and 8.5× reductions in area and power, respectively.
ER -