arXiv:2610.38922
Adaptive Local Measurements For Quantum Learning Can Exponentially Outperform Nonadaptive Global Measurements
Alireza Goldar, Zhen Qin, Zhihui Zhu, Zhe-Xuan Gong, Michael B. Wakin
Abstract
Can measurements chosen adaptively, in response to earlier outcomes, be exponentially more informative than measurements fixed in advance? In many quantum settings, adaptivity helps at most polynomially; known exponential advantages require coherent measurements across copies or restricted measurements. We show an exponential advantage measuring one copy at a time, with no quantum memory kept between state copies, even when the nonadaptive learner may perform arbitrary global measurements on each copy. We construct explicit families of multi-qubit states in which a scalar parameter is encoded in one of exponentially many candidate observables, with an auxiliary marker indicating which one. An adaptive protocol identifies the relevant observable and then measures it using only single-qubit measurements and one round of adaptation. At fixed accuracy, it needs logarithmically many copies as the qubit count grows, whereas every nonadaptive protocol needs exponentially many. The same mechanism yields an exponential gap in local asymptotic minimax mean squared error constants, tight up to the exponential base. The states are separable and classical in label-dependent product bases; no entanglement or other nonclassical correlations are required. The gap comes from measurement incompatibility: the candidate observables cannot be measured jointly, so a fixed measurement must divide its information budget among them, while adaptivity selects the informative one. Accordingly, the separation disappears when the candidate observables commute, and cannot occur with fully observed classical samples. Measurement incompatibility thus emerges as a resource for quantum learning, distinct from entanglement and quantum memory, that classical feedback alone can exploit.
AI Summary
Can classical feedback give an exponential advantage in quantum learning without collective measurements across copies? This paper constructs a scalar-estimation problem where adaptive single-qubit measurements outperform every nonadaptive protocol allowing arbitrary joint POVMs within each copy. Each $n=2L$-qubit state has a marker encoding an unknown word $\mathbf a\in\{X,Y,Z\}^L$ and a signal $\sigma_{\mathbf a,c}=2^{-L}(I+cP_{\mathbf a})$, where $P_{\mathbf a}=\bigotimes_\ell a_\ell$. Markers distinguish local labels through biases $-u,0,u$, with $0<u<1/2$. One adaptation round first identifies the word, then measures its Pauli parity. At fixed $u$ and accuracy $\epsilon$, this uses $O(u^{-2}\log L+\epsilon^{-2})$ copies, while nonadaptive global measurements require at least $1/(144B_L\epsilon^2)$ copies for $0<\epsilon\leq1/4$, where $B_L=(2^L-1)(1+u)^L/3^L$. The lower bound follows from an aggregate Fisher-information budget, obtained using the Gill–Massar inequality and a bound on marker side information, then converted to two-point testing bounds. For local asymptotic minimax MSE around $c=0$, the adaptive $1/N$ constant is $1$, whereas the nonadaptive constant is at least $B_L^{-1}$; a 2-design measurement gives an upper bound $2^L+1$. Every state is separable, diagonal in a label-dependent product basis, and has MPO bond dimension at most two under marker–signal blocking. A commuting binary construction loses its advantage under global measurements, identifying measurement incompatibility as the operative obstruction. Collective cross-copy measurements are excluded; the exact exponential base, stronger markers, neighborhoods away from zero, and robustness to model mismatch remain unresolved.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 32 pages, 0 figures