A hybrid quantum-classical neural network with deep residual learning.

Area of Science:

• Quantum Computing
• Artificial Intelligence
• Machine Learning

Background:

• Classical neural networks have shown significant success.
• Efforts are underway to translate these successes into the quantum computing domain.
• Deep residual learning is a powerful technique in classical deep learning.

Purpose of the Study:

• To propose a novel hybrid quantum-classical neural network incorporating deep residual learning.
• To analyze the integration of residual block structures within quantum neural networks.
• To develop an end-to-end training algorithm analogous to backpropagation.

Main Methods:

• Introduction of the Residual Hybrid Quantum-Classical Neural Network (Res-HQCNN).
• Analysis of connecting residual block structures with quantum neural networks.
• Development of a training algorithm for end-to-end learning.
• Experimental validation on classical computers using quantum data, including noisy scenarios.

Main Results:

• The Res-HQCNN model can be trained in an end-to-end fashion.
• Res-HQCNN outperforms existing methods in learning unknown unitary transformations.
• The proposed network exhibits enhanced robustness when dealing with noisy quantum data.

Conclusions:

• The Res-HQCNN offers a promising approach for hybrid quantum-classical machine learning.
• Deep residual learning can be effectively integrated with quantum neural networks.
• The model shows potential for advancing quantum data analysis and processing.