One major goal of the current generation of quantum computers is to “simulate” (or emulate) the Hamiltonians found in condensed matter and material systems. Such quantum simulation strategies are particularly important in cases where it is challenging to simulate these systems with traditional computational tools, such as quantum Monte Carlo or tensor network methods - numerical schemes that have been under development for decades. Recently, the rapidly-advancing field of machine learning has introduced a host of new methods suitable for this task, involving neural network architectures and data-driven learning strategies. In this talk, I will discuss the complementary role of experimental and in silico quantum simulations through the lens of machine learning, using the example of present-day Rydberg atom quantum computers. In particular, I will illustrate the utility of machine learning methods to leverage data from real experiments, and speculate on the future of scientific discovery in quantum many-body simulators that hybridize traditional and data-driven approaches.
