SAMOP 2023 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

QI: Fachverband Quanteninformation

QI 23: Poster II (joint session QI/Q)

QI 23.2: Poster

Wednesday, March 8, 2023, 16:30–19:00, Empore Lichthof

Introducing Non-Linear Activations into Quantum Generative Models — •Mykolas Sveistrys^{1, 2}, Kaitlin Gili², and Chris Ballance² — ¹Dahlem Center for Complex Quantum Systems, Freie Universität Berlin, Arnimallee 14, 14195 Berlin, Germany — ²Department of Physics, University of Oxford, Clarendon Laboratory, Parks Road, Oxford OX1 3PU, U.K.

One prominent difference between most classical generative models and current quantum ones is linearity: classical neural-network-based models require non-linear activations for quality training, while embedding such activations in quantum models is challenging due to the linearity of quantum mechanics. We introduce a quantum generative model that adds non-linear activations via a neural network structure onto the standard Born Machine framework - the Quantum Neuron Born Machine (QNBM). We utilize a previously introduced Quantum Neuron subroutine, which is a repeat-until-success circuit with mid-circuit measurements and classical control. We then compare the QNBM to the linear Quantum Circuit Born Machine (QCBM). With gradient-based training, we show that while both models can easily learn a trivial uniform probability distribution, on a more challenging class of distributions, the QNBM achieves an almost 3x smaller error rate than a QCBM with a similar number of tunable parameters. We therefore provide evidence that suggests that non-linearity is a useful resource in quantum generative models, and we put forth the QNBM as a new model with good generative performance and potential for quantum advantage.