arXiv:2610.19544
Shot- and Noise-Aware Pauli Correlation Encoding for Max-Cut Experiments on IBM Heron Quantum Processor
Achraf Boussahi
BOUSSAHI ACHRAFAbstract
Pauli correlation encoding (PCE) encodes $m$ binary variables in the signs of $k$-body Pauli correlators of $n=O(m^{1/k})$ qubits, and trains a parameterized circuit on a relaxed loss of sharpness $α$. The strongest published schedules sharpen this loss aggressively and are benchmarked with exact expectation values, whereas a quantum computer estimates correlators from a finite number of shots. We show that the sharpness is a hidden shot budget. Since the expected sign of a correlator estimated from $S$ shots, $\mathrm{erf}(c\sqrt{S/2})$, agrees with $\tanh(0.851\sqrt{S}\,c)$ to within $0.019$, the edge term of the relaxed loss at $α=0.851\sqrt{S}$ approximately equals, up to a constant, minus twice the cut expected from decoding $S$ shots. This explains why schedules tuned with exact expectation values lose their advantage at realistic budgets, and yields a shot-aware schedule whose final sharpness is set by the shot budget. On five G-set graphs with $800$ variables on $13$ qubits, this schedule outperforms every published schedule at its target budget in all $150$ paired comparisons of the decoded circuit output, owing to the sharpness at $1.6\times10^4$ shots and mostly to switching off the regularization at $10^3$; after a classical bit-swap search, it remains the best on average at $1.6\times10^4$ shots. Device noise damps the correlators by a factor $s$, which amounts to a smaller shot budget $s^2S$. On the IBM Heron processor ibm_fez, an all-to-all ansatz of the type used for the best published G-set results compiles to $835$ CZ gates and decodes to a random cut, while our $96$-CZ hardware-native circuit, trained in simulation and sharpened to $α\simeq0.65\,s\sqrt{S}$, retains up to $0.919$ of the best-known cut. For new circuits, a damping and an offset calibrated on other circuits of the same graph predicted the decoded cuts at $10^3$ shots to within $0.002$.
AI Summary
Pauli correlation encoding (PCE) stores binary variables in signs of Pauli expectation values and trains circuits using a $\tanh(\alpha c)$ relaxation. This paper asks how its sharpness $\alpha$ should depend on finite-shot readout and device noise, rather than exact-expectation benchmarks. Using a Gaussian approximation for small correlators and neglecting correlations between decoded signs, it relates the expected decoded spin, $\mathrm{erf}(c\sqrt{S/2})$, to $\tanh(0.851\sqrt{S}\,c)$, with uniform sigmoid discrepancy below $0.019$. Consequently, the loss’s edge term approximately optimizes the expected cut at an implicit budget $S=(\alpha/0.851)^2$. Multiplicative correlator damping $s$ reduces this to an effective budget $s^2S$, motivating a budget-dependent sharpening schedule with final regularization removed. Exact-simulation training on five 800-variable G-set graphs encoded in 13 qubits compares schedules across five paired initializations. The proposed schedule wins all 150 comparisons against three published schedules at its target budgets, $10^3$ and $1.6\times10^4$ shots per setting. Classical bit-swap post-processing erases most gains at the lower budget. IBM Heron executions compare a 96-CZ native circuit with an 835-CZ all-to-all stand-in: the latter yields random-cut performance, while native circuits reach up to $0.919$ of the best-known cut. Sharpness sweeps are consistent with an empirical optimum $\alpha\simeq0.65s\sqrt{S}$. All training uses exact simulation; hardware tests cover single trained circuits per schedule, not hardware training. These sizes remain classically simulable and are not competitive with classical solvers. Damping calibration uses known ideal correlators; scalable calibration remains untested. Uneven hardware damping limits predictive accuracy, especially at higher shot budgets.
Error found: The abstract's loss–expected-cut relation is established only conditionally on Gaussian, small-correlator, and negligible decoded-spin-dependence approximations, not generally for arbitrary PCE circuits. The expected-sign formula omits finite-shot ties under sgn(0)=+1, giving zero at c=0 although the exact expectation at S=1000 is approximately 0.025225. The single effective shot budget s²S is an approximate noise model for the measured hardware, whose damping varies between settings and correlators.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 32 pages, 9 figures, 16 tables. Code and data: https://doi.org/10.5281/zenodo.23356288