Search papers, labs, and topics across Lattice.
This paper introduces a distance-4 bivariate bicycle (BB) quantum error-correction technique specifically designed for quantum convolutional neural networks (QCNNs) to mitigate the effects of noise prevalent in current quantum devices. The authors demonstrate through simulations that traditional unprotected QCNNs struggle with convergence and learning rates under realistic noise conditions, highlighting the urgent need for effective error correction. Their approach not only reduces qubit overhead but also enhances the practical viability of QCNNs, marking a significant advancement in quantum computing applications for classification tasks.
A novel error-correction method for QCNNs reduces qubit overhead while significantly improving learning rates in noisy environments.
Quantum convolutional neural networks (QCNNs) combine the power of quantum computing and classical CNN for computational speedup in classification tasks. However, noise levels on state-of-the-art quantum devices remain too high for practical QCNN execution. In addition, despite the reliable surface code providing a method for error rates below a threshold value, they have a prohibitively large qubit cost. Recently introduced bivariate bicycle (BB) codes are of particular interest for their high error threshold, constant encoding rate, and linear code distance. Through simulation with realistic hardware noise sources, we demonstrate that a 4-qubit unprotected QCNN fails to converge and exhibits a worse learning rate compared to numerical simulations. Addressing both limitations, we propose a distance-4 BB quantum error-correction (QEC) technique for QCNNs. In doing so, we validate that our low-overhead QEC technique for QCNNS represents a step toward practical QCNNs.
