Digital-Analog Quantum Machine Learning

Abstract

Machine Learning algorithms are extensively used in an increasing number of systems, applications, technologies, and products, both in industry and in society as a whole. They enable computing devices to learn from previous experience and therefore improve their performance in a certain context or environment. In this way, many useful possibilities have been made accesible. However, dealing with an increasing amount of data poses difficulties for classical devices. Quantum systems may offer a way forward, possibly enabling to scale up machine learning calculations in certain contexts. On the other hand, quantum systems themselves are also hard to scale up, due to decoherence and the fragility of quantum superpositions. In the short and mid term, it has been evidenced that a quantum paradigm that combines evolution under large analog blocks with discrete quantum gates, may be fruitful to achieve new knowledge of classical and quantum systems with no need of having a fault-tolerant quantum computer. In this Perspective, we review some recent works that employ this digital-analog quantum paradigm to carry out efficient machine learning calculations with current quantum devices.