Distributed quantum machine learning via classical communication

TL;DR

This paper introduces a practical distributed quantum machine learning approach that uses classical communication to connect quantum processors, improving data classification accuracy and enabling scalable quantum learning on intermediate-scale hardware.

Contribution

It presents an experimentally feasible scheme for distributed quantum machine learning utilizing classical communication, demonstrated with quantum convolutional neural networks on synthetic datasets.

Findings

Classical communication enhances classification accuracy.

Accuracy with classical communication matches that with quantum communication at tested depths.

The scheme is implementable with current quantum hardware capabilities.

Abstract

Quantum machine learning is emerging as a promising application of quantum computing due to its distinct way of encoding and processing data. It is believed that large-scale quantum machine learning demonstrates substantial advantages over classical counterparts, but a reliable scale-up is hindered by the fragile nature of quantum systems. Here we present an experimentally accessible distributed quantum machine learning scheme that integrates quantum processor units via classical communication. As a demonstration, we perform data classification tasks on 8-dimensional synthetic datasets by emulating two 4-qubit processors and employing quantum convolutional neural networks. Our results indicate that incorporating classical communication notably improves classification accuracy compared to schemes without communication. Furthermore, at the tested circuit depths, we observe that the accuracy with classical communication is no less than that achieved with quantum communication. Our work provides a practical path to demonstrating large-scale quantum machine learning on intermediate-scale quantum processors by leveraging classical communication that can be implemented through currently available mid-circuit measurements.