en.Wedoany.com Reported - WiMi Hologram Cloud Inc. (WiMi) has released core technology for hybrid quantum convolutional neural networks, proposing and implementing a quantum kernel convolution (QKC) scheme suitable for current noisy intermediate-scale quantum (NISQ) devices, providing a feasible path for the engineering of quantum-enhanced image classification models.
The core objective of this technology is not to directly embed quantum circuits into classical neural networks, but to rethink the computational methods of feature extraction and dimensionality reduction starting from computationally intensive convolution operations. WiMi explains that classical convolutional layers rely on sliding windows and linear weighted summation for local feature extraction, while quantum computing offers high-dimensional Hilbert space representation and quantum parallelism. If local image patches can be mapped to quantum states and feature mixing achieved through controlled entanglement evolution, it may be possible to obtain equivalent or even more expressive feature extraction capabilities with fewer parameters.
WiMi points out that its proposed pooling method is essentially an information redistribution and selection mechanism, enabling dimensionality compression without explicitly discarding information, thereby reducing the computational burden on subsequent quantum circuits and classical networks.
In terms of overall system architecture, this hybrid QCNN adopts a classical-quantum collaborative hierarchical design. The classical neural network is responsible for initial normalization, dimensionality adjustment, and final classification decisions of input data, while the quantum convolutional layer is embedded at key feature extraction positions, functioning as a quantum acceleration module. This design allows the model to fully leverage mature classical deep learning toolchains while introducing quantum advantages at critical computational nodes, avoiding scalability issues of fully quantum models under current hardware conditions.
In terms of technical implementation, WiMi has completed a full engineering implementation based on the Qiskit quantum computing development framework, from quantum circuit construction and parameterized training to integration with classical deep learning frameworks. The quantum convolutional layer is encapsulated as a reusable module interface that can be directly embedded into existing deep learning training workflows. A hybrid optimization strategy is adopted during training: classical backpropagation algorithms update classical network parameters, while parameter shift rules estimate quantum circuit gradients, enabling end-to-end joint training. This approach addresses the gradient propagation challenge between quantum and classical components.
During the experimental phase, WiMi selected the MNIST handwritten digit dataset as a benchmark task to systematically evaluate the hybrid QCNN model. Results show that, with a significant reduction in parameter count compared to traditional CNN models, the hybrid model still achieves classification accuracy comparable to classical models. Notably, after replacing some classical convolutional layers with quantum convolutional layers, the model's parameter scale and computational complexity are effectively controlled while maintaining stable convergence performance, demonstrating the practical feasibility of quantum kernel convolution in real-world tasks.
Through analysis of intermediate quantum states and measurement results, WiMi verified the effectiveness of the entanglement-based quantum pooling mechanism in the dimensionality reduction process. Experiments show that quantum pooling not only compresses feature dimensions but also retains key discriminative information required for classification tasks. This finding provides a new entry point for research on the interpretability of quantum neural networks and lays the foundation for scaling to more complex datasets and tasks.
This hybrid quantum convolutional neural network technology represents a significant step forward in WiMi's long-term strategic goal of deployable quantum-enhanced artificial intelligence. By emphasizing low depth, modularity, and high compatibility with existing AI ecosystems, this technology offers a realistic path for quantum computing to move from the laboratory to practical applications. WiMi stated that it will further explore the application potential of this architecture in higher-resolution images, multi-channel data, and other perception tasks, while continuously optimizing circuit design in conjunction with quantum hardware development.
The release of WiMi's hybrid neural network quantum kernel convolution technology marks an important step for quantum machine learning from concept validation to engineering implementation, demonstrating the practical value of quantum computing in real-world image recognition tasks and providing a clear design approach for building quantum-classical collaborative computing systems. As quantum hardware performance continues to improve and development tools mature, WiMi's hybrid QCNN framework is expected to play a role in a wider range of AI applications, becoming an important component of next-generation intelligent computing technology.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
