WiMi Explores Hybrid Quantum-Classical Convolutional Neural Network Model

BEIJING, Oct. 23, 2025 — WiMi Hologram Cloud Inc. (NASDAQ: WiMi) (“WiMi” or the “Company”), a prominent global provider of Hologram Augmented Reality (“AR”) Technology, announced today that it is actively exploring a shallow hybrid quantum-classical convolutional neural network (SHQCNN) model, which promises innovative advancements in the domain of image classification.
Variational quantum methods, recognized as crucial technical tools in quantum computing, offer effective avenues for designing and implementing quantum algorithms by converting quantum state optimization challenges into classical optimization problems. WiMi has incorporated an enhanced variational quantum method into its SHQCNN model, thereby establishing a robust foundation for the model’s efficient execution in image classification tasks. This refined variational quantum method includes several key optimizations over conventional techniques. Firstly, in terms of quantum state representation, it more accurately describes the quantum characteristics of image data through the introduction of more complex combinations of quantum gates and parameterized forms. Secondly, within the optimization algorithm, sophisticated adaptive optimization strategies are employed. These strategies dynamically adjust optimization parameters based on real-time feedback during the training phase, accelerating convergence and boosting the model’s training efficiency. This improved variational quantum method allows the SHQCNN model to fully capitalize on quantum computing’s strengths when handling image classification tasks, while simultaneously circumventing the complexity issues that typically arise from increasing layers in traditional Quantum Neural Networks (QNNs).
In image classification, the model’s performance is directly influenced by the quality and distinctiveness of the input data. The SHQCNN model integrates a kernel encoding method in its input layer, which serves as a precise mechanism to enhance data differentiation and processing efficiency. The fundamental concept of kernel encoding involves mapping original image data from a low-dimensional space to a high-dimensional feature space using nonlinear transformations. This process makes image data that is difficult to distinguish in lower dimensions much easier to separate in higher dimensions. By utilizing the kernel encoding method, the SHQCNN model optimizes data processing at the input stage, ensuring high-quality input for the computations performed by subsequent hidden and output layers, thereby boosting the overall classification accuracy of the model.
Furthermore, the hidden layer, being a core component of the neural network, carries out the vital function of extracting and transforming input data features. In traditional QNNs, an increase in the number of layers leads to a sharp rise in the hidden layer’s computational complexity, making the training process exceptionally difficult. The SHQCNN model cleverly addresses this challenge by designing variational quantum circuits within its hidden layer. Variational quantum circuits are constructed from a series of quantum gates, capable of performing specific transformations on the input quantum states. Compared to the hidden layers found in traditional deep neural networks, variational quantum circuits exhibit a more concise structure and lower computational complexity. Through judicious design of quantum gate types and arrangement order, these circuits can efficiently extract image features with fewer layers. Concurrently, the parameters of the variational quantum circuit can be optimized using classical algorithms, enabling the model to adaptively fine-tune itself for various image classification tasks, thus further enhancing its generalization capability.
The output layer, functioning as the neural network’s final module, is responsible for making classification decisions based on the features extracted by the hidden layer. The SHQCNN model incorporates the mini-batch gradient descent algorithm in its output layer; this innovative application of the algorithm significantly improves the model’s parameter training and learning speed. The mini-batch gradient descent algorithm is a variant of the standard gradient descent algorithm. In each iteration, instead of processing the entire training dataset, it randomly selects a small subset (“mini-batch”) of data for computation. This approach results in faster computation and superior convergence compared to the traditional batch gradient descent algorithm. Within the SHQCNN model, the mini-batch gradient descent algorithm allows for more frequent weight updates, enabling timely adjustments to the model’s parameters and accelerating its adaptation to changes in the training data.
The shallow hybrid quantum-classical convolutional neural network model (SHQCNN) developed by WiMi, through its integrated application of advanced technologies such as enhanced variational quantum methods, kernel encoding, variational quantum circuits, and mini-batch gradient descent algorithms, demonstrates significant advantages in stability, accuracy, and generalization. This model is set to introduce novel solutions to the field of image classification. With the ongoing advancements in quantum computing technology and the broadening scope of its application scenarios, the SHQCNN model is expected to unlock its vast potential across an even wider range of domains.

About WiMi Hologram Cloud
WiMi Hologram Cloud Inc. (NASDAQ: WiMi) specializes in holographic cloud services, primarily focusing on professional areas such as in-vehicle AR holographic Head-Up Displays (HUD), 3D holographic pulse LiDAR, light-field holographic head-mounted devices, holographic semiconductors, holographic cloud software, holographic automotive navigation, metaverse holographic AR/VR devices, and metaverse holographic cloud software. The company encompasses a broad spectrum of holographic AR technologies, including in-vehicle holographic AR, 3D holographic pulse LiDAR, holographic vision semiconductors, holographic software development, holographic AR virtual advertising, holographic AR virtual entertainment, holographic ARSDK payment solutions, interactive holographic virtual communication, metaverse holographic AR, and metaverse virtual cloud services. WiMi positions itself as a holistic provider of holographic cloud technology solutions. For more details, please visit .
Translation Disclaimer
The initial version of this announcement is the officially authorized and only legally binding document. Should any discrepancies or semantic differences arise between the Chinese translation and the original text, the original version will take precedence. WiMi Hologram Cloud Inc., along with its associated entities and individuals, offers no assurances regarding the translated version and disclaims all liability for any direct or indirect losses resulting from translation inaccuracies.
For Investor Inquiries, please reach out to:
WIMI Hologram Cloud Inc.
Email: pr@wimiar.com

ICR, LLC
Robin Yang
Tel: +1 (646) 975-9495
Email: