Coherent consensus for the estimation and classification of phases

Abstract

Inspired by quantum metrology, we propose a coherent consensus primitive to enhance the collaborative performance of independent quantum classifiers at the decision boundary. The basic idea consists of three steps: i) prepare a Greenberger-Horne-Zeilinger (GHZ) state on $n$ qubits; ii) apply a relative phase to each qubit, which encodes the output of one classifier; and iii) undo the GHZ preparation to collect all phases onto the first qubit. The signal is amplified by a factor of $n$, which reduces the sampling error by a factor of $\sqrt{n}$ relative to incoherent majority voting, that is, the direct combination of estimates from individual classifiers. Coherent consensus is valid for general estimation and classification of phases. We derive guarantees on the resources needed to estimate phases using this coherent consensus primitive as compared to majority voting. We implement both techniques on a trapped-ion quantum computer using a given ensemble of imperfect phase gates. We observe the expected $\sqrt{n}$ gain in precision of coherent consensus, with excess dispersion at the largest $n$ consistent with device drift. Coherent consensus can thus be understood as a form of sensing inside a quantum computer.

Abstract: arXiv descriptive metadata (CC0). Thank you to arXiv for use of its open access interoperability. This service was not reviewed or approved by, nor does it necessarily express or reflect the policies or opinions of, arXiv.