Speaker
Description
One of the core challenges of research in quantum computing is to understand whether quantum advantages can be found for problems of practical interest. In the field of quantum machine learning, we know for a few years now that proof-of-concept exponential advantages can be established in learning tasks derived from classically hard problems such as factoring or computing discrete logarithms. In this talk, I will expose recent efforts in making these learning separations more physically and practically relevant. Namely, I will present a first result establishing how non-local correlations can be at the origin of learning advantages in the experimentally-friendly realm of shallow-depth circuits. I will also explain how computationally-universal quantum many-body dynamics can be efficiently learned on a quantum computer, while remaining classically intractable under standard complexity assumptions.