At the time of writing, quantum machine learning (QML) is just about the greatest combination of buzzwords you could hope to synthesize. A lot is written about QML, and the topic is often (confusingly) both overhyped and undersold at the same time. In this section we’ll try to give a flavor for how QPUs might transform machine learning, while also being careful to point out the caveats inherent in manipulating quantum data.
Useful QML applications require very large numbers of qubits. For this reason, our overview of QML applications is necessarily very high-level. Such a summary is also fitting given the rapidly changing nature of this nascent field. Although our discussion will be more schematic than pragmatic, it will heavily leverage our hands-on experience of primitives from earlier chapters.
We summarize three different QML applications: solving systems of linear equations, Quantum Principal Component Analysis, and Quantum Support Vector Machines. These have been selected due to both their relevance to machine learning and their simplicity to discuss. These are also applications whose conventional counterparts are hopefully familiar to anyone who has dabbled in machine learning. We only give a brief description of the conventional progenitors of each QML application as it’s introduced.
In discussing QML, we’ll frequently make use of the following pieces of machine-learning terminology:
Term used to describe measurable properties ...