Speaker: Alejandro Pozas
Affiliation: UCM
Date: Tuesday, 7 June 2022 at 12:00
Location: Online seminar
Physics research has not been alien to the recent success of machine learning techniques. Unlike other disciplines, physics is in a unique position to influence research in machine learning as well. In this talk, I will argue and practically illustrate that insights in quantum information, concretely coming from the tensor network representations of quantum many-body states, can help in devising better privacy-preserving machine learning algorithms. After a short introduction to privacy in machine learning, I will show that standard neural networks are vulnerable to a type of privacy leak that, a priori, is resistant to standard protection mechanisms. Afterwards, I will show that tensor networks, when used as machine learning architectures, are invulnerable to this leak. The proof of resilience is based on the existence of canonical forms for such architectures. Given the growing expertise in training tensor networks and that these architectures are recently showing to compete and even surpass traditional machine learning architectures, these results imply that one may not have to be forced to make a choice between accuracy in prediction and ensuring the privacy of the information processed when using machine learning on sensitive data. This talk is based on arXiv:2202.12319.