WiMi Hologram Cloud Inc. reports developments across holographic augmented-reality technology, quantum computing research and semiconductor-related services. The company describes itself as a holographic cloud technology solution provider, with work in AR advertising and entertainment, in-vehicle AR HUD software, 3D holographic pulse LiDAR, head-mounted light-field devices, holographic semiconductors, metaverse AR/VR devices and related cloud software.
Recurring WIMI news centers on releases of hybrid quantum-classical neural networks, quantum convolutional models, feature-mapping methods, text and image classification technologies, and applications of quantum modules in machine-learning architectures. Company updates also include annual Form 20-F announcements, operating expense and profitability commentary, and subsidiary-related capital developments.
WiMi (NASDAQ: WIMI) proposed a hybrid quantum-classical Inception neural network for image classification that integrates three parallel paths—quantum, classical, and hybrid—to improve performance, efficiency, and robustness. The design uses shallow, high-entanglement quantum circuits, parameterized rotation encoding, and concatenated multi-path features to enhance expressiveness while reducing parameter counts.
The architecture targets better trainability and scalability by replacing deep quantum circuits with parallel shallow circuits and plans future work toward hardware deployment and deeper hybrid structures.
WiMi (NASDAQ:WIMI) released a Hybrid Quantum-Classical Neural Network (H-QNN) for efficient MNIST binary image classification on Feb 6, 2026. The H-QNN uses a parameterized quantum circuit for feature encoding, quantum-state feature extraction, and a classical MLP classifier with hybrid optimization. WiMi reports ~30% lower simulation compute time versus comparable classical networks and observed improved feature expressivity when scaling qubits from 4 to 8.
The framework targets extensibility to handwriting recognition, medical imaging, and video-frame feature extraction, and WiMi plans device-level verification and broader quantum-algorithm integration.
WiMi (NASDAQ: WIMI) announced a new Lean Classical-Quantum Hybrid Neural Network (LCQHNN) designed to improve image classification while minimizing quantum resource use. The architecture pairs a lightweight classical front-end for feature extraction with a four-layer variational quantum circuit (4-layer VQC) back-end for nonlinear mapping and classification. WiMi says amplitude/phase encoding, entanglement via controlled rotations and CNOTs, and a parameter-shift training method reduce measurement counts and hardware error accumulation. The company plans multimodal extensions, prototype hardware deployment, and federated quantum optimization.
WiMi (NASDAQ: WIMI) on January 5, 2026 announced MC-QCNN, a next-generation Quantum Convolutional Neural Network for Multi-Channel Supervised Learning. The company says the design creates hardware-adaptable quantum convolution kernels that encode multi-channel data into amplitudes, phases, or entanglement and use parameterized gates, SWAP interleaving, weak entanglement, and learnable quantum pooling to preserve features.
WiMi describes a hybrid quantum-classical training framework, extended parameter-shift training, noise simulation, and claimed advantages for image classification, medical imaging, video analysis, and multimodal monitoring.
WiMi (NASDAQ: WIMI) announced QB-Net, a hybrid quantum-classical deep learning approach that embeds a pluggable Quantum Bottleneck Module into the U-Net architecture. WiMi says QB-Net reduces the bottleneck layer parameter count by up to 30x while maintaining performance comparable to classical U-Net.
QB-Net encodes classical features into quantum states, applies parameterized quantum circuits with entanglement for feature transformation, then decodes measurements back into classical tensors. The module is designed for minimal parameters, trainability, and plug-and-play integration without changing U-Net structure or training paradigms.
WiMi (NASDAQ: WIMI) announced a next‑generation hybrid quantum neural network (H-QNN) for image multi‑classification on Dec 22, 2025. The H-QNN combines classical convolutional neural networks for spatial feature extraction with quantum neural networks for high‑dimensional nonlinear mapping, using a three‑stage design: feature dimensionality reduction & encoding, quantum state transformation, and hybrid decision & transfer learning. Key technical elements include PCA plus angle encoding, parameter sharing to mitigate barren plateaus, an early stopping strategy using quantum Fidelity, and an FPGA‑accelerated quantum simulation module claiming nanosecond‑level state updates and superior training speed versus pure CPU/GPU simulations.
The design supports simulation and hardware QPU execution and emphasizes transfer learning to reduce epochs and improve stability in multi‑class tasks.
WiMi (NASDAQ: WIMI) announced a hybrid quantum-classical learning architecture for multi-class image classification on December 4, 2025. The design reuses measurement results from both retained and discarded qubits in quantum convolutional neural networks (QCNN), feeding them into separate classical fully connected branches that are fused and jointly trained with quantum parameters.
The approach aims to reduce information loss from quantum pooling on NISQ devices, improve expressive power of hybrid models, and enable co-evolution of quantum gate angles and classical weights via cross-entropy backpropagation. WiMi positions this as a practical path for QML under current hardware constraints, with applications in intelligent vision, medical diagnosis, and autonomous driving.
WiMi (NASDAQ: WIMI) announced a research architecture for quantum generative adversarial networks (QGANs) on Nov 20, 2025, proposing a dual-discriminator hybrid quantum-classical design built around a quantum convolutional neural network (QCNN) discriminator.
The design uses a three-layer QCNN discriminator pipeline—quantum feature encoding, parallel quantum feature extraction, and classical decision output—to improve feature capture, shorten gradient paths, and reduce gradient vanishing via particle swarm optimization of quantum gate parameters. Two QCNN discriminators are trained with dynamically balanced loss weights to force simultaneous global distribution matching and local feature authenticity, aiming to boost stability and diversity of QGAN outputs for tasks like image generation and quantum state simulation.
WiMi (NASDAQ:WIMI) announced research into a quantum generative adversarial network (QGAN) and a hybrid quantum-classical convolutional neural network for synthetic image generation and detection on November 14, 2025.
The company says the QGAN yields shorter simulation and training time and lower generator and discriminator losses versus traditional GANs, and the hybrid model pairs quantum computing with classical CNNs to improve feature extraction and classification accuracy for real vs. synthetic images.
The release frames these models as research-stage advancements that could expand applications as quantum hardware and algorithms improve.
WiMi Hologram Cloud (NASDAQ: WIMI) announced on October 24, 2025 that it is researching a blockchain privacy protection system built on post‑quantum cryptography combined with threshold key‑sharing. The design targets quantum‑resistant encryption, distributed private‑key management across consensus nodes, threshold‑based data authorization, fine‑grained access control, and support for smart contracts and scalable consensus.
The system emphasizes post‑quantum security, fault tolerance via distributed key shares, controlled data access through node thresholds, and optimizations for high concurrency and dynamic scalability.