Beijing, Jan. 15, 2026 (GLOBE NEWSWIRE) -- WiMi Studies Quantum Hybrid Neural Network Model to Empower Intelligent Image Classification
BEIJING, Jan. 15, 2026––WiMi Hologram Cloud Inc. (NASDAQ: WiMi) ("WiMi" or the "Company"), a leading global Hologram Augmented Reality ("AR") Technology provider, proposed a brand-new Lean Classical-Quantum Hybrid Neural Network (LCQHNN) framework, aimed at achieving maximized learning efficiency with the smallest possible quantum circuit structure. This technology balances implementability and performance superiority in its design, marking a key step for quantum neural networks from theoretical feasibility toward practical deployment.
The core idea of LCQHNN is to center on quantum feature amplification (Quantum Feature Amplification) while combining a classical stability optimization strategy, establishing an efficient information interaction mechanism between the two computing paradigms. The network architecture is divided into two main parts:
Classical Front-End: responsible for preliminary feature extraction and data pre-encoding;
Quantum Back-End: utilizes variational quantum circuits to complete nonlinear mapping and classification decisions.
In this system, the classical part uses lightweight convolutional and fully connected layers as the data preprocessing channel, with their output results embedded into the quantum state space and subjected to feature transformation through parameterized quantum gate operations. This process is equivalent to mapping high-dimensional classical features to a multi-dimensional quantum Hilbert space, thereby forming nonlinear projections in superposition states, enabling the model to capture the essence of complex data distributions with fewer parameters.
In the quantum part, WiMi designed a structure containing only a four-layer variational quantum circuit (4-layer VQC). This circuit consists of parameterized rotation gates, controlled gates, and entanglement operations. Through optimization of the circuit parameters, the relationship between the measurement results of the quantum state output and the target categories gradually converges. Experiments show that a four-layer circuit can achieve performance comparable to or even better than deep VQCs, thereby significantly reducing the resource consumption and error accumulation risk of quantum hardware.
The complete workflow of WiMi's LCQHNN can be summarized into the following key stages:
Data Preprocessing and Classical Encoding: The original image first undergoes lightweight convolutional layers to extract local features, followed by normalization and compression operations to form a medium-dimensional vector representation. Subsequently, these vectors are mapped into input states encoded by quantum amplitudes or phases. For example, amplitude encoding can compress high-dimensional data into a limited number of qubits, allowing classical information to be stored in the quantum state space in an exponential manner.
Quantum State Preparation and Entanglement Structure Construction: After encoding is completed, the system enters the quantum section. WiMi employs controlled rotation gates and CNOT gates to construct entanglement structures, enhancing correlations between different qubits. The introduction of this entanglement pattern not only improves the expressive power of the quantum feature space but also theoretically endows the model with stronger nonlinear discrimination capability. Research results show that an appropriate number of entanglement layers is one of the key determining factors for model performance, and in LCQHNN, the four-layer variational structure design precisely balances performance and implementability.
Parameterized Quantum Evolution and Measurable Readout: Each layer of the quantum circuit contains adjustable parameters θ, which correspond to the angles of rotation gates or phase shift gates. Through multiple evolutions and measurements of the quantum state, the system collects the statistical distribution of measurement results, thereby constructing a loss function that can be used for gradient backpropagation. WiMi adopts an improved gradient estimation method—an efficient training mechanism based on the parameter shift rule—which significantly reduces the number of quantum measurements required for each parameter update, improving overall training speed and stability.
Classical Feedback and Hybrid Optimization: During the optimization process, the backpropagation algorithm of the classical part runs in coordination with the parameter updates of the quantum part. Classical optimizers (such as Adam or L-BFGS) are responsible for adjusting the quantum circuit parameters θ so that the measurement results minimize classification error. This process embodies the core concept of hybrid quantum-classical collaborative optimization: fully leveraging the high-dimensional expressive power of the quantum feature space while building on the stability of classical computation.
Classification Decision and Feature Visualization: The final quantum measurement results are decoded back into the classical domain and used to output the category to which the image belongs. Through characterization analysis, WiMi found that LCQHNN can form distinct feature cluster distributions during training. These clusters correspond to different quantum state distribution regions in quantum space, exhibiting strong inter-class separability.
The success of LCQHNN has laid a solid foundation for constructing a General Quantum Intelligence Framework. In the future, the research team plans to continue expanding in the following directions: extending the model to multimodal learning scenarios to achieve joint quantum feature learning for images, speech, and text; exploring collaborative integration with quantum support vector machines (QSVM) and quantum convolutional networks (QCNN) to build end-to-end quantum deep learning systems; promoting prototype deployment on quantum hardware to verify the model's performance stability in real noisy environments; and combining quantum parallel optimization with federated learning frameworks to construct secure, efficient, and distributed quantum intelligent systems.
The launch of WiMi's Lean Classical-Quantum Hybrid Neural Network (LCQHNN) marks a new stage in which quantum machine learning technology has moved from theoretical exploration toward efficient practical implementation. By achieving outstanding learning performance under limited quantum resources, this technology not only makes breakthrough progress in image classification tasks but also provides a new paradigm for the design of future quantum intelligent systems. WiMi will continue to devote itself to the engineering and industrialization promotion of quantum algorithms, driving quantum artificial intelligence from the laboratory to real-world application scenarios and accelerating humanity's entry into the era of quantum intelligence.
About WiMi Hologram Cloud
WiMi Hologram Cloud Inc. (NASDAQ: WiMi) focuses on holographic cloud services, primarily concentrating on professional fields such as in-vehicle AR holographic HUD, 3D holographic pulse LiDAR, head-mounted light field holographic devices, holographic semiconductors, holographic cloud software, holographic car navigation, metaverse holographic AR/VR devices, and metaverse holographic cloud software. It covers multiple aspects of holographic AR technologies, including in-vehicle holographic AR technology, 3D holographic pulse LiDAR technology, holographic vision semiconductor technology, holographic software development, holographic AR virtual advertising technology, holographic AR virtual entertainment technology, holographic ARSDK payment, interactive holographic virtual communication, metaverse holographic AR technology, and metaverse virtual cloud services. WiMi is a comprehensive holographic cloud technology solution provider. For more information, please visit http://ir.wimiar.com.
Translation Disclaimer
The original version of this announcement is the officially authorized and only legally binding version. If there are any inconsistencies or differences in meaning between the Chinese translation and the original version, the original version shall prevail. WiMi Hologram Cloud Inc. and related institutions and individuals make no guarantees regarding the translated version and assume no responsibility for any direct or indirect losses caused by translation inaccuracies.
Investor Inquiries, please contact:
WIMI Hologram Cloud Inc.
Email: pr@wimiar.com
ICR, LLC
Robin Yang
Tel: +1 (646) 975-9495
Email: wimi@icrinc.com
