Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper formalizes the mismatch between translation equivariance in classical CNNs and existing Quantum CNNs (QCNNs) that use address/amplitude encodings like FRQI. It then introduces Pixel Cyclic Shift (PCS)-equivariant QCNN layers built using quantum Fourier transforms (QFTs) to diagonalize translations. The authors demonstrate a deep PCS-QCNN architecture with measurement-induced pooling and prove a lower bound on the expected squared gradient norm, suggesting improved trainability compared to other deep quantum networks.
QCNNs can now achieve exact pixel-translation equivariance, just like classical CNNs, by leveraging Fourier multiplexers, potentially unlocking more effective quantum machine learning.
Convolutional neural networks owe much of their success to hard-coding translation equivariance. Quantum convolutional neural networks (QCNNs) have been proposed as near-term quantum analogues, but the relevant notion of translation depends on the data encoding. For address/amplitude encodings such as FRQI, a pixel shift acts as modular addition on an index register, whereas many MERA-inspired QCNNs are equivariant only under cyclic permutations of physical qubits. We formalize this mismatch and construct QCNN layers that commute exactly with the pixel cyclic shift (PCS) symmetry induced by the encoding. Our main technical result is a constructive characterization of all PCS-equivariant unitaries: conjugation by the quantum Fourier transform (QFT) diagonalizes translations, so any PCS-equivariant layer is a Fourier-mode multiplexer followed by an inverse QFT (IQFT). Building on this characterization, we introduce a deep PCS-QCNN with measurement-induced pooling, deferred conditioning, and inter-layer QFT cancellation. We also analyze trainability at random initialization and prove a lower bound on the expected squared gradient norm that remains constant in a depth-scaling regime, ruling out a depth-induced barren plateau in that sense.