Developing Quantum Reservoir Computing as Machine Learning of Music

This paper introduces a Quantum Reservoir Computing approach to learning time-based sequences of events. It focuses on music as a domain application example. Both the learning process of music and the theory of reservoir computing aligning with each other are explored, validating the suitability and potential of quantum reservoirs in excelling at such tasks. Initial experiments were conducted with a musical piece hinting at the exponential increase in efficiency relative to conventional AI models. The paper is organised as follows: It begins with an introduction to the tasks of high-dimensional analysis and temporal learning, which are required for music. Then, it introduces the basics of Reservoir Computing more generally and discusses how it can address those analysis and learning problems. Next, it shows how concepts from Quantum Computing can be leveraged to harness Reservoir Computing, leading to the concept of Quantum Reservoir Computing. Then, it shows the use of a conventional AI model to learn music in comparison with a classical Reservoir Computing approach. Finally, it presents our approach to using Quantum Reservoir Computing to learn and generate music with an example. The paper concludes with final remarks on the results and challenges for future work.