Quantum CNNs achieve translation equivariance via Fourier multiplexers

Researchers have developed a new type of quantum convolutional neural network (QCNN) that is designed to be equivariant to pixel shifts, a property crucial for image recognition tasks. This advancement addresses a mismatch in existing QCNNs where translation equivariance is often limited to cyclic permutations of qubits. The proposed QCNN utilizes Fourier transforms to achieve exact pixel cyclic shift equivariance, potentially improving performance on quantum data processing tasks. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Introduces a new QCNN architecture that could improve image processing capabilities on quantum computers.

RANK_REASON This is a research paper detailing a novel QCNN architecture.

Read on arXiv cs.LG →

• arXiv
• Quantum Convolutional Neural Networks
• Fourier Multiplexers
• Quantum Fourier Transform
• Dmitry Chirkov

• paper
• other