Quantum reservoir computing (QRC) is a quantum machine learning (QML) algorithm that uses a quantum reservoir to process datasets and extract information which is later fed to a classicalmachine learning model.The reservoir could be implemented as a quantum circuit in a noisy intermediate-scale quantum(NISQ) computer. A recently developed criterion based on the majorization principle can be appliedto select optimal quantum reservoirs, rendering better results than other common models withsignificantly less gates.The presence of noise difficults QRC, correcting or mitigating the induced errors is costly. But, can webenefit from noise? Surprisingly, we will show that under some specific circumstances, quantum noisecan be used to improve the performance of QRC. Certain noise types can be beneficial to machinelearning, while others should be prioritized for correction. This gives practical prescriptions forsuccessful implementations in nowadays hardware.
