arXiv:2610.28070
Measurement incompatibility in Bayesian quantum estimation beyond quadratic loss
Francesco Albarelli, Stefano Chessa, Jesús Rubio
Abstract
In single-shot Bayesian multiparameter quantum estimation, measurement incompatibility can increase the minimum joint quadratic loss by at most a factor of two relative to the bound obtained by optimising each parameter separately. Using the generalised Barnum--Knill inequality, we establish the same bound for a wider class of additive losses generated by positive score functions, showing that this limitation extends beyond the standard quadratic setting. To illustrate this result, we consider a broad family of periodic losses, including the sine loss sometimes employed in interferometry, and derive the corresponding incompatibility bound for Bayesian multiphase estimation with arbitrary prior knowledge. We further extend the framework to unbounded losses, such as absolute losses; this also recovers the case of quadratic losses.
AI Summary
In single-shot Bayesian multiparameter estimation, measurement incompatibility increases the optimal additive quadratic loss by at most a factor of two relative to the sum of separately optimized parameter losses. This paper asks whether that bound survives for nonquadratic losses, particularly periodic costs used in phase estimation. For additive losses $\ell=\sum_j w_j\ell_j$, the benchmark is $\mathcal L_{\mathrm{IF}}=\sum_j w_j\mathcal L_j^*$, where each single-parameter optimization treats the other unknown parameters as nuisance parameters. If $\ell_j=c_j(1-\sigma_j)$ with $\sigma_j\in[0,1]$ admitting a positive-kernel factorization, the paper proves $\mathcal L_{\mathrm{IF}}\leq\mathcal L_{\mathrm{opt}}\leq\mathcal L(M_{\mathrm{PGM}})\leq2\mathcal L_{\mathrm{IF}}$. The same prior-dependent pretty-good measurement achieves this guarantee simultaneously for every component. The proof applies the generalized Barnum–Knill inequality to each component’s score and converts score guarantees into loss bounds. Applications include additive periodic Holevo losses with nonpositive nonconstant Fourier coefficients, including sine loss, for arbitrary priors. For sine loss, the benchmark is expressed through trace norms of first circular-moment operators. An extension using positive exponential kernels $e^{-\kappa\ell_j}$ and monotone convergence covers unbounded losses, including absolute and quadratic losses; conditionally negative-definite losses satisfy this kernel condition. The results require additive losses and the stated positivity conditions; they do not establish the bound for general losses, including tolerance-window scores. Infinite-dimensional extensions are anticipated but their regularity assumptions are not worked out. The benchmark isolates measurement incompatibility rather than the additional penalty from unknown nuisance parameters.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 7 pages, no figures