The recoverable quantum information: A smart reward for reinforcement learning of quantum error correction

Quantum information can be encoded in a complex manner in entangled multi-qubit states. But how well is this information preserved, given the dissipative time evolution of a system? In order to address this question, we have developed a new quantity, which we call the recoverable quantum information.

Beyond the possibility to judge how well existing techniques perform, it is particularly useful for the autonomous discovery of new quantum error correction strategies, by using it as an immediate reward scheme for a reinforcement learning approach as demonstrated in [1]. The recoverable quantum information is applicable generically, i.e. independent of the concrete hardware platform, noise model, etc. In this talk, this quantity will be introduced and motivated, an intuitive interpretation is given, it is shown how it evolves in some exemplary systems, and I will sketch how it can be computed numerically.

[1] T. Fösel, P. Tighineanu, T. Weiss, and F. Marquardt, Phys. Rev. X 8, 031084 (2018)