Quantum Photonic Architectures and Machine Learning

Quantum devices and quantum algorithms promise substantial advantages in many computational tasks, with a significant speedup when compared to classical computers on specific tasks. In recent years, an ever-increasing number of advancements is being made in this field, with the first experiments tackling the quantum advantage regime, namely the scenario where quantum devices are capable of outperforming classical computers in specific tasks. These results have thus opened the way for the development and application of noisy intermediate-scale quantum (NISQ) processors for quantum-enhanced information processing. At the same time, machine learning has been finding use in many different sectors, due to its capabilities and its flexibility. The meeting point between these two fields, Quantum Machine Learning (QML), has thus started garnering interest, as a valid approach to devise quantum algorithms capable of performing well even on modern noisy quantum devices. However, one of the challenges that quantum machine learning is facing is how to efficiently train these QML models, since some techniques that are used for training classical machine learning models cannot be used in the quantum case. In this thesis we will address quantum machine learning in different frameworks, going from validation of currently available hardware, to training quantum machine learning algorithms and proposing a new quantum optical architecture. The main focus of this thesis will be on Quantum Machine Learning on the photonic platform, including how to efficiently train such quantum machine learning models, showing their capabilities when properly trained, and proposing a novel quantum optical architecture, which makes use of an optical component obtained by training an optical quantum machine learning model. Thus, the objective of this thesis is to address how to properly train these QML algorithms, as well as highlighting relevant use cases where these quantum machine learning models can provide an advantage when compared to other techniques.