Learning a quantum channel from its steady-state

TL;DR

This paper introduces a scalable method for learning non-unital quantum channels by analyzing their steady states, enabling efficient characterization of noise models and quantum processes using local measurements.

Contribution

The authors develop a novel approach to learn non-unital quantum channels from steady-state expectation values, extending techniques from Hamiltonian learning to quantum channels.

Findings

Method successfully learns quantum channels from steady states.

Applicable to noise model assessment and parameter estimation.

Validated through simulations and IBMQ experiments.

Abstract

We present a scalable method for learning local quantum channels using local expectation values measured on a single state -- their steady state. Our method is inspired by the algorithms for learning local Hamiltonians from their ground states. For it to succeed, the steady state must be non-trivial, and therefore the channel needs to be non-unital. Such non-unital channels are readily implementable on present day quantum computers using mid-circuit measurements or RESET gates. We demonstrate that the full structure of such channels is encoded in their steady states, and can be learned efficiently using only the expectation values of local observables on these states. We emphasize two immediate applications to illustrate our approach: (i) Using engineered dissipative dynamics, we offer a straightforward way to assess the accuracy of a given noise model in a regime where all qubits are actively utilized for a significant duration. (ii) Given a parameterized noise model for the entire system, our method can learn its underlying parameters. We demonstrate both applications using numerical simulations and experimental trials conducted on an IBMQ machine.