Relatore: Andrea Mari (Xanadu Quantum Technologies)
Abstract
Despite the impressive experimental progress of the last years, all current quantum computers are still relatively small and noisy. For this reason, in the near future, they are not expected to be practically useful for implementing many standard quantum information algorithms (e.g. Shor’s factorization). On the other hand, quantum processors have already reached a regime in which they can perform computations which are classically intractable. In this scenario, hybrid quantum machine learning models composed of quantum and classical computational layers are very promising. In this talk I will introduce a new paradigm for designing and training hybrid neural networks. I will also present some numerical simulations performed with the quantum software library PennyLane and a proof-of-concept example in which high-resolution images have been successfully classified with real quantum computers (by Rigetti and IBM).