arXiv:2609.30374
Quantum advantage in learning single mode bosonic channels
Angela Anna Baiju, Aritra Das, Özlem Erkılıç, Jiayi Qin, Li Gong, Syed M Assad, Ping Koy Lam, Biveen Shajilal, Lorcán O Conlon, Jie Zhao
Abstract
Quantum resources can dramatically reduce the data required to learn physical systems, with exponential improvements in sample complexity demonstrated in several quantum learning tasks. However, these advantages have typically relied on quantum resources that scale with problem complexity, most notably increasing system dimension or entanglement. This raises a fundamental question: can exponential quantum learning advantages arise within a fixed, unentangled quantum system? In this paper, we address this question by considering the learning of an unknown random-displacement distribution whose complexity is determined not by the dimensionality of the physical system, but by the Fourier resolution of its features. We show that the quantum-limited noise of vacuum probes progressively obscures high-frequency features, leading to an exponential growth in sample complexity. We establish an information-theoretic lower bound for arbitrary classical-state probes and show that squeezing overcomes this classical limit by extending the accessible Fourier bandwidth. Experimentally, we demonstrate an exponential reduction in sample complexity using squeezed vacuum probes for both binary hypothesis testing and characteristic-function reconstruction. Our results show that a single bosonic mode can exhibit exponential quantum learning advantages without entanglement or an increase in system size, identifying accessible Fourier bandwidth as a fundamentally distinct resource for quantum-enhanced learning.
AI Summary
The paper asks whether exponential sample-complexity advantages can arise without entanglement or increasing the number of bosonic modes. It studies a single-mode, single-quadrature random-displacement channel, distinguishing a Gaussian displacement distribution from the same Gaussian modulated by $1+4\epsilon_0\sin(2\beta_0x)$. Increasing $\beta_0$ introduces finer structure without changing the second moment. For Gaussian probes followed by homodyne detection, quantum noise convolves the displacement distribution. An unbiased characteristic-function estimator has variance bounded by $N^{-1}e^{4a\beta^2}$, where $a=1$ for vacuum and $a=e^{-2r}$ for ideal squeezed vacuum. A log-likelihood test achieves discrimination complexity proportional to $\epsilon_0^{-2}\exp[4a\beta_0^2/(1+4a\sigma_\beta^2)]$. A sandwiched-Rényi-divergence calculation establishes an arbitrary-collective-measurement lower bound for parallel positive-$P$ classical probes: $N=\Omega(\exp[\beta_0^2\sigma_\beta^{-2}(1-(1+4\sigma_\beta^2)^{-1/2})])$. Sufficient fixed squeezing beats this bound exponentially; ideal squeezing growing as $r=\ln(2\beta_0)$ gives constant-sample discrimination. An optical implementation uses 4.95–8.54 dB squeezing and homodyne detection. At $\beta_0=4.5$, discrimination requires approximately $10^4$ squeezed-probe samples versus $10^7$ vacuum-probe samples. The reported advantage over the arbitrary-measurement classical lower bound is approximately tenfold. Reconstruction comparisons reach eight orders of magnitude, with large vacuum requirements inferred beyond available data. The setting is restricted to structured one-quadrature distributions. Experimental reconstruction tests 25 frequencies rather than continuous uniform accuracy; the displayed Hoeffding derivation controls an endpoint component. Loss imposes a noise floor, and noise-calibration errors are exponentially amplified at high frequencies. Constant-sample scaling assumes increasing ideal squeezing, not fixed energy.
Error found: The paper does not establish its continuum-uniform reconstruction upper bound or the corresponding constant-sample corollary, because its concentration calculation controls only one component at one frequency. The claim that the classical lower bound covers arbitrary adaptive channel-use strategies is broader than the supplied proof, which covers parallel classical inputs and collective output measurements. The reported eight-order reconstruction advantage compares squeezed-probe measurements with an unmeasured vacuum baseline at the quoted bandwidth, despite being described as directly demonstrated experimentally.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 9 pages, comments and suggestions very welcome