Statistical learning is emerging as a new paradigm in science.
This has ignited interest within our inherently quantum world in exploring quantum machines for their advantages in learning, generating, and predicting various aspects of our universe by processing both quantum and classical data. In parallel, the pursuit of scalable science through physical simulations using both digital and analog quantum computers is rising on the horizon.
In the first part, we investigate how physics can help classical Artificial Intelligence (AI) by studying hybrid classical-quantum algorithms. We focus on quantum generative models and address challenges like barren plateaus during the training of quantum machines. We further examine the generalization capabilities of quantum machine learning models, phase transitions in the over-parameterized regime using random matrix theory, and their effective behavior approximated by Gaussian processes.
In the second part, we explore how AI can benefit physics. We demonstrate how classical Machine Learning (ML) models can assist in state recognition in qubit systems within solid-state devices. Additionally, we show how ML-inspired optimization methods can enhance the efficiency of digital quantum simulations with ion-trap setups
Finally, in the third part, we focus on how physics can help physics by using quantum systems to simulate other quantum systems. We propose native fermionic analog quantum systems with fermion-spin systems in silicon to explore non-perturbative phenomena in quantum field theory, offering early applications for lattice gauge theory models.
*We strongly encourage attendees to use their full name (and if possible, their UMD credentials) to join the zoom session.*