2019/01/03 by Jonathan Romero, Alán Aspuru‐Guzik, Romero, Jonathan +1 · 13 citations
Computer Science · #FOS: Physical sciences #Neural Networks and Applications #Quantum Computing Algorithms and Architecture #Quantum Information and Cryptography #Quantum Physics (quant-ph)
paper · pdf · doi:10.48550/arxiv.1901.00848
openalex publication_date 2019/01/03 · openalex created_date 2022/07/30 · openalex updated_date 2026/08/01
We propose a hybrid quantum-classical approach to model continuous classical\nprobability distributions using a variational quantum circuit. The architecture\nof the variational circuit consists of two parts: a quantum circuit employed to\nencode a classical random variable into a quantum state, called the quantum\nencoder, and a variational circuit whose parameters are optimized to mimic a\ntarget probability distribution. Samples are generated by measuring the\nexpectation values of a set of operators chosen at the beginning of the\ncalculation. Our quantum generator can be complemented with a classical\nfunction, such as a neural network, as part of the classical post-processing.\nWe demonstrate the application of the quantum variational generator using a\ngenerative adversarial learning approach, where the quantum generator is\ntrained via its interaction with a discriminator model that compares the\ngenerated samples with those coming from the real data distribution. We show\nthat our quantum generator is able to learn target probability distributions\nusing either a classical neural network or a variational quantum circuit as the\ndiscriminator. Our implementation takes advantage of automatic differentiation\ntools to perform the optimization of the variational circuits employed. The\nframework presented here for the design and implementation of variational\nquantum generators can serve as a blueprint for designing hybrid\nquantum-classical architectures for other machine learning tasks on near-term\nquantum devices.\n