TY - GEN

T1 - Simulation of a Small-Scale Chemical Reservoir Computer for Pattern Recognition

AU - Kopets, Ekaterina

AU - Tatiana, Shchetinina

AU - Rybin, Vyacheslav

AU - Dautov, Albert

AU - Karimov, Timur

AU - Karimov, Artur

N1 - Conference code: 10.1109/MECO55406.2022

PY - 2022/6/21

Y1 - 2022/6/21

N2 - Reservoir computing is a framework within which an arbitrary non-trainable reservoir, e.g. a network with a fixed structure, is used for computing, and the output of the reservoir is supplied with a simple classification or regression algorithm, which is trained to yield required results. In the current study, we investigate the possibility of a simple spiking neural network (SNN) based on the extended Brusselator model to serve as the reservoir for binary pattern recognition. We train various classifiers on data obtained from the four-neuron SNN and show that when there are two and three digits in a binary pattern three and six unique patterns can be distinguished, respectively, with the positive predictive value not less than 89 %. The obtained results can be used for developing a fast-training chemical computer taking advantage of its hardware structure.

AB - Reservoir computing is a framework within which an arbitrary non-trainable reservoir, e.g. a network with a fixed structure, is used for computing, and the output of the reservoir is supplied with a simple classification or regression algorithm, which is trained to yield required results. In the current study, we investigate the possibility of a simple spiking neural network (SNN) based on the extended Brusselator model to serve as the reservoir for binary pattern recognition. We train various classifiers on data obtained from the four-neuron SNN and show that when there are two and three digits in a binary pattern three and six unique patterns can be distinguished, respectively, with the positive predictive value not less than 89 %. The obtained results can be used for developing a fast-training chemical computer taking advantage of its hardware structure.

KW - Brusselator

KW - SNN

KW - clas-sification learner

KW - pattern recognition

KW - reservoir computer

KW - spiking neural networks

UR - https://www.mendeley.com/catalogue/25fad9f8-bb5b-39c4-bffa-b30f4c44713e/

U2 - 10.1109/MECO55406.2022.9797166

DO - 10.1109/MECO55406.2022.9797166

M3 - статья в сборнике материалов конференции

SN - 9781665468282

BT - 2022 11th Mediterranean Conference on Embedded Computing, MECO 2022

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - Mediterranean Conference on Embedded Computing

Y2 - 7 June 2022 through 10 June 2022

ER -