A PAC-Bayesian account in which quantum policy design is reframe around an entanglement--generalization trade-off rather than expressivity alone is given, where entangled circuits consistently generalize worse than non-entangled circuits of equal parameter count.
Abstract
Parameterized quantum circuits (PQCs) are increasingly used as policies and value functions in quantum reinforcement learning, yet it remains unclear when and why quantum policies generalize. We give a PAC-Bayesian account in which generalization is governed not by the raw number of circuit parameters, but by the effective dimension of the Fisher geometry induced by the circuit. This quantity is inflated by entanglement, making entangling connectivity an independent axis of complexity.In controlled experiments that fix the number of trainable rotations and vary only entanglement, we find that circuits with larger Fisher effective dimension exhibit larger train-test gaps, while parameter count is a weak predictor. The resulting bound acts primarily as a ranking certificate: it correctly orders circuits with identical parameter count, which parameter-counting bounds cannot do. We validate this mechanism across supervised classification, quantum contextual bandits, and value-function generalization, where entangled circuits consistently generalize worse than non-entangled circuits of equal parameter count, with gaps shrinking as sample size increases.Our strongest evidence comes from low-variance decision models, including single-observable classifiers, value heads, and one-step policies. In end-to-end multi-step policy learning, entanglement effects remain statistically significant but high return variance leaves the full ordering only partially resolved. Partial-correlation analysis shows that Fisher effective dimension screens off entangling pattern, and controls for training accuracy, readout, and optimizer rule out major optimization confounders. The effect also persists on an IBM Heron quantum processor under real noise. Overall, our results reframe quantum policy design around an entanglement--generalization trade-off rather than expressivity alone.
This paper provides a detailed comparison of the performance of separable and entangled Quantum Physics-Informed Neural Networks (QPINNs) to solve coupled nonlinear differential equations, with the multi-generator power system swing equation serving as a test application. Previous studies focused on single-layer circuits not enforcing initial conditions, while the work here implements multi-layer variational circuits that include explicit initial condition loss. All models are tested against both a high-accuracy RK45 numerical baseline and iso-parameter classical PINNs. An ablation study considering qubit count and circuit depth shows that entanglement is the most important architectural property because it provides approximately 10× lower physics residuals than separable QPINNs at equivalent parameter budgets for entangled QPINNs. An additional study of performance based on the number of generators shows the differential between performance of entangled QPINNs and classical PINNs becomes smaller as the system grows larger (40 machines for 2 machines and 27 machines for 4 machines). This supports the theory that a ring CNOT topology replicates physical connection between generators. This research demonstrates that entangled QPINNs are suitable candidates for power system stability analysis on NISQ hardware with 280× fewer parameters than classical PINNs with the same architecture.
L. Kavisankar, Sayak Das, Prashoon Mishra· 2026 International Conferenc...· 0 citations
While entanglement is known to enable exponential improvements in the sample complexity of quantum learning, it remains unclear which properties of entangled resources are responsible for such improvements. We address this question through the reduction criterion, a condition obeyed by all bound-entangled states. In $n$-qubit Pauli-channel learning, we show that restricting either the input states or the measurement effects to satisfy this criterion rules out an exponential advantage for incoherent adaptive protocols. An exponential lower bound persists for the one-sided coherent adaptive protocols considered here, even when the unrestricted side retains quantum correlations across channel uses. Using conditional min-entropy, we further quantify how the sample-complexity lower bounds weaken as larger violations of the reduction criterion are allowed. Finally, we show that the same obstruction appears in conjugate-state learning: restricted joint measurements cannot reproduce the logarithmic-sample advantage of unrestricted joint measurements on $\rho\otimes\rho^*$. These results identify violation of the reduction criterion as a necessary condition for an exponential advantage in the learning tasks considered here.
This work constructs a native 2D pairwise ansatz and compares its expressibility and trainability with representative 1D ansatze at identical layer depths, despite their different circuit depths.
Quantum state ensembles are important in quantum information processing. For example, quantum $t$-designs model highly entangled states in complex systems, while projected ensembles appear in generative quantum machine learning and studies of thermalization. With their sample state accompanied by a classical label, these ensembles contain operational information beyond their average density operators. Yet an ensemble differs from a classical-quantum state because it is invariant under permutations of labels. We formulate binary hypothesis testing between finite quantum ensembles and derive fundamental limits on error probability. Given an observed label pattern, we show that the joint sampled state can be described by power-weighted ensemble moments. This yields the Bayes-optimal measurement and exact finite-sample error, revealing that discrimination is governed by the full moment hierarchy up to the number of samples. In the many-sample limit, we derive Chernoff bounds and obtain exact error exponents for finite uniform pure-state ensembles. We apply these results to optical communication and $t$-designs. For finite uniform pure-state $t$-designs with large $t$, the maximal discrimination exponent scales sharply as $\sim t^{-2}$, while equal-prior fixed-error testing requires $\sim t^2$ samples.
Programmable quantum devices nominally act on a Hilbert space whose dimension grows exponentially with the number of constituents, but the presence of noise makes it unlikely that they remain coherent across all of this immense Hilbert space. Then, what is the effective coherent quantum dimension that should be associated with such imperfect devices? To answer this question we here introduce an operational basis-independent framework which imposes a dimension bottleneck on the programmable transformations. Concretely we ask how strongly the quantum information they process can be compressed. Formalizing this idea we identify three inequivalent notions, termed $d$-compressibility, $d$-simulability and $d$-embeddability, which differ in the causal structure used to impose the bottleneck and form a strict hierarchy. The framework unifies several existing notions: joint measurability and simulability of quantum measurements, and the absolute dimensionality of state ensembles, are recovered as special cases. We illustrate the hierarchy with noisy qubit measurements in complementary bases, and we determine the white-noise thresholds at which the set of all noisy unitary channels in dimension $n$, a noisy universal quantum processor, becomes $d$-compressible, $d$-simulable and $d$-embeddable. The thresholds confirm the expectation -- maintaining coherence across the full Hilbert space becomes increasingly demanding as the nominal dimension $n$ increases.
It is shown that gradient-based PQCs can exhibit improved performance on unseen data as model size increases, displaying the phenomenon of double descent, which contrasts with the traditional view that larger models lead to degraded generalization.
Marie C. Kempkes, Elies Gil-Fuster, Carlos Bravo-Prieto et al.· 0 citations