Multiclass classiﬁcation using quantum convolutional neural networks with hybrid quantum-classical learning

Bokhan, D.; Mastiukova, A.S.; Boev, A.S.; Trubnikov, D.N.; Fedorov, A.K.

Авторы: Bokhan Denis, Mastiukova Alena S., Boev Aleksey S., Trubnikov D.N., Fedorov Aleksey K.
Журнал: FRONTIERS IN PHYSICS
Том: 10
Год издания: 2022
Первая страница: 1069985:1
Последняя страница: 1069985:9
Номер статьи: 1069985
DOI: 10.3389/fphy.2022.1069985
Аннотация: Multiclass classiﬁcation is of great interest for various applications, for example,it is a common task in computer vision, where one needs to categorize an imageinto three or more classes. Here we propose a quantum machine learningapproach based on quantum convolutional neural networks for solving themulticlass classiﬁcation problem. The corresponding learning procedure isimplemented via TensorFlowQuantum as a hybrid quantum-classical(variational) model, where quantum output results are fed to the softmaxactivation function with the subsequent minimization of the cross entropyloss via optimizing the parameters of the quantum circuit. Our conceptionalimprovements here include a new model for a quantum perceptron and anoptimized structure of the quantum circuit. We use the proposed approach tosolve a 4-class classiﬁcation problem for the case of the MNIST dataset usingeight qubits for data encoding and four ancilla qubits; previous results havebeen obtained for 3-class classiﬁcation problems. Our results show thataccuracies of our solution are similar to classical convolutional neuralnetworks with comparable numbers of trainable parameters. We expect thatour ﬁnding provide a new step towards the use of quantum neural networks forsolving relevant problems in the NISQ era and beyond.
Добавил в систему: Бохан Денис Александрович

	ИСТИНА	Войти в систему Регистрация
	Интеллектуальная Система Тематического Исследования НАукометрических данных
	Главная Поиск Статистика О проекте Помощь

ИСТИНА

Интеллектуальная Система Тематического Исследования НАукометрических данных

Multiclass classiﬁcation using quantum convolutional neural networks with hybrid quantum-classical learningстатья

Прикрепленные файлы