WiMi Launches Two-Qubit QCNN Model Matching Traditional CNN Accuracy with Fewer Parameters
WIMI•WiMi Hologram Cloud completed benchmark testing of a fully parameterized Quantum Convolutional Neural Network model that uses only two-qubit interactions to simplify circuit depth and reduce noise accumulation. The model matches or exceeds traditional CNN classification accuracy with fewer parameters, positioning WiMi for future deployment on intermediate-scale quantum hardware.
1. Completed Benchmark Testing
WiMi’s research team systematically benchmarked the fully parameterized QCNN model on classical image classification tasks, demonstrating stable convergence, effective noise control and consistent accuracy across diverse test scenarios.
2. Simplified Two-Qubit Architecture
The quantum convolutional network relies exclusively on two-qubit interactions as its base computational unit, effectively controlling circuit depth and mitigating noise accumulation to enable practical implementation on current noisy intermediate-scale quantum computers.
3. Data Preprocessing and Encoding
Raw images undergo dimensionality reduction via principal component analysis, feature selection and compression before entering the quantum encoding stage; researchers compared angle encoding, amplitude encoding and hybrid methods to balance representation capacity with qubit resource constraints.
4. Superior Classification Performance
Experimental results reveal that the QCNN matches or exceeds the classification accuracy of traditional CNNs while using significantly fewer trainable parameters, highlighting enhanced parameter utilization efficiency enabled by quantum entanglement in high-dimensional feature spaces.
