arXiv:2610.10339
Binary Optimization of Measurement Groupings for Quantum Energy Estimation
Isaac L. Huidobro-Meezs, Rodrigo A. Vargas-Hernández
Abstract
Repeated measurements can dominate the resources required for quantum energy estimation, making the choice of which Pauli observables to measure together a central optimization problem for variational quantum algorithms. We formulate fully commuting measurement grouping as a classical binary optimization problem based on clique selection and construct non-overlapping groups using mixed-integer linear programming (MILP). For a benchmark of molecular Hamiltonians, groupings optimized with approximate covariances reduce the non-overlapping measurement requirement $\varepsilon^2M$ by $51.8\%$ on average relative to sorted insertion (SI). Using the MILP groups to initialize iterative coefficient splitting (ICS), denoted MILP-ICS, yields an average $24.3\%$ reduction relative to ICS initialized from SI (SI-ICS). The optimized groups also transfer across nearby molecular geometries while preserving substantial measurement savings. We further introduce O-clique, which directly optimizes overlapping commuting supports through candidate-clique selection and coefficient profiles. Although O-clique provides only modest additional reductions beyond MILP-ICS, it reduces the measurement requirement by $27.3\%$ on average relative to SI-ICS with 100 iterations, despite using only five final coefficient refinement iterations, indicating improved support quality. Finally, we extend the comparison to Fermi--Hubbard, Kitaev--Heisenberg--$Γ$, and XYZ lattice Hamiltonians, where MILP-based groupings substantially outperform the corresponding SI-based strategies. Together, these results show that variance-informed optimization of measurement-group structure can substantially reduce sampling costs across molecular and lattice Hamiltonians, with optimized non-overlapping groups providing strong, transferable initializations and direct overlapping optimization offering further gains.
AI Summary
The paper asks how covariance-informed selection of commuting Pauli measurement groups affects energy-estimation sampling costs, both directly and as initialization for overlapping coefficient splitting. It optimizes the shot-allocation objective $\varepsilon^2M=(\sum_\alpha\sqrt{\operatorname{Var}(H_\alpha)})^2$, rather than group count. For non-overlapping groups, candidate commuting cliques receive precomputed standard-deviation costs, converting restricted partition selection into a mixed-integer linear program (MILP), supplemented by greedy refinement. The resulting partitions initialize iterative coefficient splitting (MILP-ICS). A second method, O-clique, selects overlapping supports using fixed coefficient profiles and continuous profile multipliers. Its linear surrogate $U$ satisfies $S\leq U$ and $\varepsilon^2M\leq U^2$ for the covariance data used; fixed-support ICS subsequently refines the coefficients. Across eight small molecular Hamiltonians, optimization with approximate CISD covariances, followed by evaluation with exact FCI covariances, reduces measurement requirements by 51.8% on average relative to sorted insertion (SI). MILP-ICS reduces costs by about 24% relative to SI-ICS. O-clique achieves a 27.3% average reduction relative to SI-ICS with 100 iterations, using five final refinement iterations, but adds generally less than 10% beyond MILP-ICS, with a maximum of 13%. Geometry-transfer tests retain savings, and three lattice benchmarks also favor MILP-based groupings over SI-based counterparts. Optimality certificates apply only within finite candidate pools, not globally. Classical search costs vary substantially, reported timings exclude covariance construction, and transferred groupings can degrade when new terms appear. The lattice tests use exact ground-state covariances rather than demonstrating approximate-covariance performance.
Error found: The claimed 0.5% measurement reduction from adding LNS is unreconciled with the displayed MILP and MILP-LNS tables, which show an aggregate worsening before ICS and different improvement percentages after ICS.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 18 pages, 4 figures, 6 tables