From blind quantum source separation to blind quantum process tomography

Abstract

We here extend the field of blind (i.e. unsupervised) quantum computation into two directions. On the one hand, we introduce a new class of blind quantum source separation (BQSS) methods, which perform quantum/classical data conversion by means of spin component measurements, followed by classical processing. They differ from our previous class of classical-processing BQSS methods by using extended types of measurements (three directions, possibly different for the considered two spins), which yield a more complete nonlinear mixing model. This allows us (i) to develop a new disentanglement-based separation procedure, which requires a much lower number of source values for adaptation and (ii) to restore a larger set of sources. On the other hand, these extended measurements motivate us to introduce a new research field, namely Blind Quantum Process Tomography, which may be seen both as the blind extension of its existing non-blind version and as the quantum extension of classical blind identification of mixing systems.