Feedback Ansatz for Adaptive-Feedback Quantum Metrology Training with Machine Learning

TL;DR

This paper introduces a machine learning-based ansatz for adaptive-feedback quantum metrology that significantly reduces computational complexity and demonstrates resilience against noise, enabling efficient high-partite system parameter estimation.

Contribution

The paper presents a novel feedback ansatz that simplifies the design of adaptive quantum metrology strategies and shows robustness against qubit loss and phase noise.

Findings

Reduced computational complexity from N^7 to N^4 for designing AFQM.

Achieved stable imprecision scaling for systems with over 207 particles.

Demonstrated resilience of feedback strategies against qubit loss and phase noise.

Abstract

It is challenging to construct metrology schemes which harness quantum features such as entanglement and coherence to surpass the standard quantum limit. We propose an ansatz for devising adaptive-feedback quantum metrology (AFQM) strategy which reduces greatly the searching space. Combined with the Markovian feedback assumption, the computational complexity for designing AFQM would decrease from $N^7$ to $N^4$ , for N probing systems. The feedback scheme devising via machine learning such as particle-swarm optimization and derivative evolution would thus requires much less time and produces equally well imprecision scaling. We have thus devised an AFQM for 207-partite system. The imprecision scaling would persist steadily for N > 207 when the parameter settings for 207-partite system is employed without further training. Our ansatz indicates an built-in resilience of the feedback strategy against qubit loss. The feedback strategies designed for the noiseless scenarios have been tested against the qubit loss noise and the phase fluctuation noise. Our numerical result confirms great resilience of the feedback strategies against the two kinds of noise.