Trainable unravelling for quantum state discrimination

We propose a way for partial compensation of maladaption of decoder to non-ideal quantum communication channel by means of optimal choice of unravelling of the decoding operation. No physical modification of the decoder itself is required. We show that it is sufficient to add an interface that modifies the decoder environment interaction. Tuning of this interface can be done by methods of (quantum-inspired) machine learning. We suggest a search algorithm for an optimal unravelling, as an alternative for the classical gradient descent method.