Regressions on quantum neural networks at maximal expressivity

Abstract

We analyze the expressivity of a universal deep neural network that can be organized as a series of nested qubit rotations, accomplished by adjustable data re-uploads. While the maximal expressive power increases with the depth of the network and the number of qubits, it is fundamentally bounded by the data encoding mechanism. Focusing on regression problems, we systematically investigate the expressivity limits for different measurements and architectures. The presence of entanglement, either by entangling layers or global measurements, saturate towards this bound. In these cases, entanglement leads to an enhancement of the approximation capabilities of the network compared to local readouts of the individual qubits in non-entangling networks. We attribute this enhancement to a larger survival set of Fourier harmonics when decomposing the output signal.