Exploring Models, Training Methods, and Quantum Supremacy in Machine Learning and Quantum Computing

Abstract

In this chapter, the authors will discuss some of the many models and training methods that have been developed in the field of machine learning to address this learning challenge. Models like neural networks and stochastic gradient descent have their own “go-to” training algorithms, each with their own set of supporting terminology and communities of experts. Since the specifics of gate decomposition, compilation, and error correction all depend heavily on the physical implementation of qubits and quantum gates, it has been difficult to design quantum hardware capable of running such algorithms. Therefore, the authors can only provide asymptotic estimations of total execution times. Since developing quantum hardware is so prohibitively expensive, researchers are incentivized to use terms like “superior quantum algorithms” to justify their work. This has given rise to the contentious term “quantum supremacy” to describe experiments that definitively show a difference between classical and quantum levels of computational complexity.