Data reconstruction based on quantum neural networks

TL;DR

This paper explores quantum neural networks and quantum autoencoders for data reconstruction, demonstrating their effectiveness on MNIST and comparing favorably with classical super-resolution methods.

Contribution

It introduces two novel quantum-based frameworks for data reconstruction and evaluates their performance against classical neural networks.

Findings

QNNs and QAE effectively reconstruct data from small datasets.

Quantum methods perform comparably to classical super-resolution networks.

Simulation results validate the potential of quantum approaches in data processing.

Abstract

Reconstruction of large-sized data from small-sized ones is an important problem in information science, and a typical example is the image super-resolution reconstruction in computer vision. Combining machine learning and quantum computing, quantum machine learning has shown the ability to accelerate data processing and provides new methods for information processing. In this paper, we propose two frameworks for data reconstruction based on quantum neural networks (QNNs) and quantum autoencoder (QAE). The effects of the two frameworks are evaluated by using the MNIST handwritten digits as datasets. Simulation results show that QNNs and QAE can work well for data reconstruction. We also compare our results with classical super-resolution neural networks, and the results of one QNN are very close to classical ones.