Measurement-based quantum machine learning

A quantum neural network (QNN) is an object that extends the notion of a classical neural network to quantum models for quantum data. We can create a QNN by parametrizing a quantum process and then using it to model unknown relations between quantum states. In this paper, we explore how to use measurement-based quantum computation for quantum machine learning problems and propose a universal QNN in this framework which we call the multiple-triangle ansatz (MuTA). Using the proposed QNN, we solve several tasks, including learning a universal set of gates, optimizing measurement with post-processing, learning a quantum instrument, and the classification of classical data. Finally, we discuss how to train an ansatz under the hardware constraints imposed by photonic Gottesman-Kitaev-Preskill qubits. Our work demonstrates the feasibility of using measurement-based quantum computation as a framework for quantum machine learning algorithms.