Experimental Implementation of Quantum Circuit Born Machines in Near-Term Quantum Devices

Finding valuable machine learning that could benefit from noisy intermediate-scale quantum computers is one of the leading research efforts towards the milestone of practical quantum advantage. In this talk, we will focus in the one of most challenging tasks for the machine learning community: the case of generative modeling in unsupervised machine learning. In Ref. [1], a data-driven quantum circuit learning (DDQCL) approach was proposed as a hybrid quantum-classical algorithm capable of training shallow quantum circuits to prepare desirable quantum states. This resulting quantum state is referred as a Quantum Circuit Born Machine (QCBM) [1,2] and it exploits the probabilistic nature of the Born amplitudes from the computational basis states to capture correlations in the classical training data set. This QCBM model can be used to solve unsupervised generative modeling tasks such as image generation and reconstruction. We will discuss results of experimental implementations of QCBMs via DDQCL, as well as ideas for DDQCL variants that could be useful in, for example, quantum state preparation and noise mitigation.

References:
[1] Benedetti et al. arXiv:1801.07686v1
[2] Liu and Wang. arXiv:1804.04168v1