Hybrid Quantum-inspired Resnet and Densenet for Pattern Recognition with Completeness Analysis

2024-03-09

arXiv.org

AI Summary

This work addresses the limited generalization and robustness of quantum-inspired neural networks in pattern recognition. We propose two novel hybrid architectures: quantum-inspired ResNet and DenseNet, integrating adaptive residual/dense connections with symmetric quantum circuit layers. To our knowledge, this is the first design that synergistically combines quantum-inspired layers with classical skip connections, effectively mitigating gradient explosion while theoretically guaranteeing noise resilience within a rigorous framework. Experimental results demonstrate that the models retain generalization performance comparable to purely classical counterparts under noisy data. Specifically, the dense variant achieves 3–4% higher accuracy than state-of-the-art hybrid quantum-classical CNNs on standard benchmarks. Moreover, it exhibits significantly enhanced robustness against asymmetric noise attacks—outperforming both purely classical models and existing hybrid approaches. These findings advance the principled integration of quantum-inspired components into deep learning architectures for improved reliability in real-world, noisy environments.

Technology Category

• Machine Learning: Quantum Machine Learning
• Computer Vision: Adversarial Attacks & Robustness
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Responsible Web: Measurement, analysis, and circumvention of Web censorship
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

In this paper, we propose two hybrid quantum-inspired neural networks with adaptive residual and dense connections respectively for pattern recognition. We explain the frameworks of the symmetrical circuit models in the quantum-inspired layers in our hybrid models. We also illustrate the potential superiority of our hybrid models to prevent gradient explosion owing to the quantum-inspired layers. Groups of numerical experiments on generalization power show that our hybrid models possess roughly the same level of generalization power as the pure classical models with different noisy datasets utilized. Furthermore, the comparison on generalization ability between our hybrid models and a state-of-the-art hybrid quantum-classical convolutional network demonstrates 3%-4% higher accuracy of our hybrid densely-connected model than the hybrid quantum-classical network. Simultaneously, in terms of groups of experiment on robustness, the results demonstrate that our two hybrid models outperform pure classical models notably in resistance to parameter attacks with various asymmetric noises. They also indicate the superiority of our densely-connected hybrid model over the hybrid quantum-classical network under both symmetrical and asymmetrical attacks. Furthermore, an ablation study indicate that the recognition accuracy of our two hybrid models is 2%-3% higher than that of the traditional quantum-inspired neural network without residual or dense connection. Eventually, we discuss the application scenarios of our hybrid models by analyzing their computational complexities.

Problem

• Hybrid quantum-inspired neural networks for pattern recognition
• Preventing gradient explosion with quantum-inspired layers
• Improving robustness and accuracy in noisy datasets

Innovation

• Hybrid quantum-inspired neural networks
• Adaptive residual and dense connections
• Superior generalization and robustness