Description
Quantum Reservoir Computing has emerged as a promising approach to exploit the complex dynamics of quantum systems for machine learning, offering an alternative to variational quantum algorithms that suffer from issues like Barren plateaus and expensive optimization. Quantum reservoirs are simple to optimize since they depend only on the training of a linear readout layer. The performance of the quantum reservoir is driven by the quantum dynamical system, and the readout layer is used to interpret the measurement results.
In this work, we investigate the use of quantum reservoir computing to create a generative model that can sample from learned distributions. We demonstrate that this approach can reproduce target distributions while retaining the simplicity of training that reservoir computing offers. Finally, we analyze the performance and limitations of these reservoirs with different distributions.