Towards Efficient Quantum Hybrid Diffusion Models

Abstract

In this paper, we propose a new methodology to design quantum hybrid diffusion models, derived from classical U-Nets with ResNet and Attention layers. Specifically, we propose two possible different hybridization schemes combining quantum computing's superior generalization with classical networks' modularity. In the first one, we acted at the vertex: ResNet convolutional layers are gradually replaced with variational circuits to create Quantum ResNet blocks. In the second proposed architecture, we extend the hybridization to the intermediate level of the encoder, due to its higher sensitivity in the feature extraction process. In order to conduct an in-depth analysis of the potential advantages stemming from the integration of quantum layers, images generated by quantum hybrid diffusion models are compared to those generated by classical models, and evaluated in terms of several quantitative metrics. The results demonstrate an advantage in using a hybrid quantum diffusion models, as they generally synthesize better-quality images and converges faster. Moreover, they show the additional advantage of having a lower number of parameters to train compared to the classical one, with a reduction that depends on the extent to which the vertex is hybridized.

arXiv comment23 pages, 14 figures (many with subfigures), 2 tables

References in corpus (13)

• Variational Quantum Algorithms
• Parameterized quantum circuits as machine learning models
• Transfer learning in hybrid classical-quantum neural networks
• Quantum advantage with shallow circuits
• The Expressive Power of Parameterized Quantum Circuits
• Experimental Quantum Generative Adversarial Networks for Image Generation
• Implementing a distance-based classifier with a quantum interference circuit
• Quantum advantage with noisy shallow circuits in 3D
• Hybrid Quantum-Classical Generative Adversarial Network for High Resolution Image Generation
• RGB Image Classification with Quantum Convolutional Ansaetze
• Style-based quantum generative adversarial networks for Monte Carlo events
• Expressivity of Quantum Neural Networks
• Quantum-Noise-Driven Generative Diffusion Models