MicroCloud Hologram Inc. Releases Hybrid Quantum-Classical Convolutional Neural Network, Achieving New Breakthrough in MNIST Multi-Class Classification

SHENZHEN, China, Oct. 24, 2025 (GLOBE NEWSWIRE) -- MicroCloud Hologram Inc. (NASDAQ:HOLO), ("HOLO" or the "Company"), a technology service provider, proposed a Quantum Convolutional Neural Network (QCNN) based on hybrid quantum-classical learning and successfully applied it to the multi-class classification problem on the MNIST dataset, achieving accuracy comparable to classical Convolutional Neural Networks (CNNs). This achievement not only demonstrates the practical feasibility of quantum computing in machine learning tasks but also provides a new pathway for application exploration in the subsequent NISQ (Noisy Intermediate-Scale Quantum) era.

The multi-class classification problem is one of the most common tasks in computer vision and artificial intelligence applications. Whether in image recognition, handwritten digit recognition, traffic sign detection, medical image analysis, or natural scene understanding, multi-class classification algorithms play an irreplaceable role. Classical convolutional neural networks have accumulated significant achievements in this field and have achieved near-human recognition performance on multiple benchmark datasets. However, at the same time, as the depth and width of models increase, the reliance of classical methods on computational resources continues to grow, with model training and inference requiring large-scale GPU/TPU clusters, making cost and energy consumption issues that cannot be ignored.

Quantum computing, with its exponential acceleration and high-dimensional information processing capabilities, provides a new approach to solving problems in artificial intelligence. In theory, quantum algorithms can significantly improve computational efficiency in certain problems through the advantages of superposition and parallel computing. It is against this backdrop that HOLO proposed and implemented a quantum convolutional neural network method based on hybrid quantum-classical learning, which was validated on the MNIST multi-class classification task.

HOLO's approach is based on a hybrid quantum-classical learning framework, which leverages the combination of classical optimizers and quantum circuits to harness the strengths of both. Specifically, the quantum component handles feature extraction and high-dimensional mapping tasks, while the classical component is responsible for loss function optimization and final classification prediction. In terms of architecture, HOLO proposed a novel Quantum Perceptron model and designed an optimized quantum circuit structure, enabling the quantum convolutional layer to efficiently extract data features.

In the input layer, this approach uses eight qubits for data encoding, which undertake the quantum representation of MNIST image information. Additionally, four auxiliary qubits are introduced to enhance the circuit's expressive power and nonlinear modeling capabilities. Through this design, the entire circuit can effectively map input data under a limited qubit scale, providing high-quality quantum features for subsequent classification tasks.

In the output stage, the measurement results of the quantum circuit are fed into a softmax activation function, and the classification error is calculated using the Cross-Entropy Loss function. Subsequently, the classical optimizer updates the parameters in the quantum circuit based on gradient feedback, thereby realizing the training process. This hybrid model not only fully utilizes the mature experience of classical optimization but also avoids the convergence difficulties associated with purely quantum training.

This technical implementation consists of four main steps, enabling the quantum circuit to be called and optimized like a classical neural network layer.

Data Encoding Stage: The MNIST dataset contains grayscale handwritten digit images. Each image is scaled and normalized, then mapped to eight qubits through Angle Encoding or Amplitude Encoding. This process transforms the two-dimensional pixel matrix into a quantum state, leveraging quantum superposition to represent more information.

Quantum Convolution Stage: In this stage, the quantum circuit implements a feature extraction function similar to a convolution kernel through quantum gate operations. Unlike the sliding of convolution kernels in classical CNNs, quantum convolution utilizes quantum entanglement and superposition states to achieve nonlinear feature combinations, thereby effectively mapping input data in high-dimensional spaces. The optimized circuit structure proposed by HOLO, by introducing auxiliary qubits, enhances the feature representation capability, enabling the model to better capture inter-class differences in multi-class classification tasks.

Quantum Pooling Stage: Classical convolutional neural networks typically use pooling layers to reduce feature dimensionality and computational complexity. In the quantum version, HOLO achieves information compression by measuring a part of qubits or through specific quantum gate operations. This not only reduces the consumption of qubit resources but also enhances the model's generalization ability to some extent.

Output and Optimization Stage: The measurement results of the quantum circuit form the model's output vector, which is transformed into a class probability distribution through the softmax activation function. The Cross-Entropy Loss function is used to measure the discrepancy between the predicted results and the true labels. The classical optimizer adjusts the quantum circuit parameters (e.g., rotation angles) based on this loss, thereby iteratively improving classification performance.

HOLO's quantum convolutional neural network demonstrates innovation in several aspects: First, HOLO designed a new quantum perceptron model that can more efficiently extract input features and provide stronger nonlinear mapping capabilities for the quantum convolutional layer. Second, the proposed optimized quantum circuit structure fully utilizes auxiliary qubits, enabling improved model performance under limited resources. Additionally, HOLO's hybrid quantum-classical learning framework, through the integration of softmax and cross-entropy, successfully achieves the optimization of quantum circuit parameters, addressing the convergence difficulties of purely quantum training.

HOLO's achievement lays the foundation for the application of quantum machine learning in real-world scenarios. In the future, methods based on quantum convolutional neural networks can be applied to more complex datasets and tasks. For example, in autonomous driving, quantum neural networks can assist vehicles in rapidly performing multi-class traffic sign recognition in real-time scenarios; in medical imaging, they can aid doctors in multi-class classification of lesions, thereby improving diagnostic efficiency; in fields such as financial risk control and security monitoring, quantum convolutional neural networks can also play a significant role.

From an industry perspective, HOLO's research provides a novel AI algorithm solution. By integrating quantum computing with classical learning, future enterprises can achieve significant advantages in terms of energy efficiency, parameter efficiency, and computational acceleration in model training. This hybrid model also offers a feasible path for practical implementation in the NISQ era, helping enterprises gain a competitive edge at the forefront of quantum technology and artificial intelligence integration. This achievement not only demonstrates the potential of quantum computing in artificial intelligence but also provides a theoretical and practical foundation for subsequent larger-scale experiments and applications. It is believed that, with the continuous advancement of quantum hardware and the ongoing improvement of hybrid learning frameworks, quantum convolutional neural networks will showcase their unique advantages in more scenarios in the future, pushing artificial intelligence to new heights.

About MicroCloud Hologram Inc.

MicroCloud is committed to providing leading holographic technology services to its customers worldwide. MicroCloud's holographic technology services include high-precision holographic light detection and ranging ("LiDAR") solutions, based on holographic technology, exclusive holographic LiDAR point cloud algorithms architecture design, breakthrough technical holographic imaging solutions, holographic LiDAR sensor chip design and holographic vehicle intelligent vision technology to service customers that provide reliable holographic advanced driver assistance systems ("ADAS"). MicroCloud also provides holographic digital twin technology services for customers and has built a proprietary holographic digital twin technology resource library. MicroCloud's holographic digital twin technology resource library captures shapes and objects in 3D holographic form by utilizing a combination of MicroCloud's holographic digital twin software, digital content, spatial data-driven data science, holographic digital cloud algorithm, and holographic 3D capture technology. MicroCloud focuses on the development of quantum computing and quantum holography, and plans to invest over $400 million in cutting-edge technology sectors, including Bitcoin-related blockchain development, quantum computing technology development, quantum holography development, and the development of derivatives and technologies in artificial intelligence and augmented reality (AR).

For more information, please visit http://ir.mcholo.com/

Safe Harbor Statement

This press release contains forward-looking statements as defined by the Private Securities Litigation Reform Act of 1995. Forward-looking statements include statements concerning plans, objectives, goals, strategies, future events or performance, and underlying assumptions and other statements that are other than statements of historical facts. When the Company uses words such as "may," "will," "intend," "should," "believe," "expect," "anticipate," "project," "estimate," or similar expressions that do not relate solely to historical matters, it is making forward-looking statements. Forward-looking statements are not guarantees of future performance and involve risks and uncertainties that may cause the actual results to differ materially from the Company's expectations discussed in the forward-looking statements. These statements are subject to uncertainties and risks including, but not limited to, the following: the Company's goals and strategies; the Company's future business development; product and service demand and acceptance; changes in technology; economic conditions; reputation and brand; the impact of competition and pricing; government regulations; fluctuations in general economic; financial condition and results of operations; the expected growth of the holographic industry and business conditions in China and the international markets the Company plans to serve and assumptions underlying or related to any of the foregoing and other risks contained in reports filed by the Company with the Securities and Exchange Commission ("SEC"), including the Company's most recently filed Annual Report on Form 10-K and current report on Form 6-K and its subsequent filings. For these reasons, among others, investors are cautioned not to place undue reliance upon any forward-looking statements in this press release. Additional factors are discussed in the Company's filings with the SEC, which are available for review at www.sec.gov. The Company undertakes no obligation to publicly revise these forward-looking statements to reflect events or circumstances that arise after the date hereof.

Contacts

MicroCloud Hologram Inc.

Email: [email protected]