arXiv:2609.32433
Feedback-based quantum optimization with low depth and measurement
Zi-Wen Huang, Jia-Cheng Fan, Xiao-Hui Ni, Su-Juan Qin, Xiao-Kai Hou, Wei Huang, Bing-Jie Xu, Fei Gao
Abstract
Feedback-based ALgorithm for Quantum OptimizatioN (FALQON) is a hybrid quantum-classical algorithm for solving combinatorial optimization problems, which circumvents classical parameter optimization but requires a deep quantum circuit. To reduce circuit depth, Arai et al. proposed second-order FALQON (SO-FALQON), achieving the best depth reduction among existing approaches. However, SO-FALQON brings a 2.3 times per-step measurement overhead, as it needs to calculate an additional second-order control coefficient. In this paper, inspired by the Backtracking Line Search (BLS) theory, we propose another method called BLS-FALQON, which not only reduces circuit depth to a comparable extent, but also achieves fewer measurements than SO-FALQON. Numerical simulations on max-cut problem with 8 to 20 vertices demonstrate that BLS-FALQON reduces the total measurement count by 37.7% compared to SO-FALQON, while maintaining a comparable circuit depth. Furthermore, we conduct real quantum hardware experiments on the Tianyan-176 quantum computer, which uses the zuchongzhi2 superconducting quantum processor, confirming that BLS-FALQON remains effective under real quantum hardware conditions.
AI Summary
FALQON avoids classical variational-parameter optimization but can require deep circuits. Its second-order variant, SO-FALQON, permits larger evolution steps at the cost of additional observable measurements. This paper asks whether comparable depth reduction is possible with lower measurement overhead. BLS-FALQON retains first-order commutator feedback and monitors the problem energy. When a layer increases the energy, it repeatedly rescales that layer’s control coefficient by $\tau\in(-1,0)$, reversing its sign and reducing its magnitude until the energy decreases. Experiments use $\tau=-1/4$. Theoretical arguments give finite-backtracking bounds under assumptions about higher-order errors or a quadratic energy landscape; they do not establish convergence to the global optimum. Simulations benchmark 274 unweighted 3-regular Max-Cut instances with 8–20 vertices, using 1,024 shots per measurement basis. To reach approximation ratio $0.932$, BLS-FALQON uses 37.7% fewer total measurement bases than SO-FALQON while requiring 2.5% more layers. Relative to original FALQON, the reported reductions are 88.1% in measurement bases and approximately 90.9% in layers. Hardware tests use 4-, 6-, and 8-qubit instances, ten layers, and twenty runs per algorithm. Estimated execution time is about 43% lower than SO-FALQON. Hardware measurements supply feedback, but reported approximation ratios are evaluated by exact simulation using the resulting controls, rather than directly from hardware output states. The paper reports increasing discrepancies from ideal trajectories with problem size and little improvement beyond roughly five layers; hardware tests do not reach the target approximation ratio.
Error found: The hardware study evaluates approximation ratios through ideal simulations with hardware-derived parameters, so it does not establish end-to-end noisy-hardware optimization performance. The finite-termination proof assumes a positive descent neighborhood that its remainder estimate and the condition A_k != 0 do not establish. Algorithm 3 retains a feedback coefficient computed from the uncorrected state after backtracking and never checks the final layer for an energy increase. Proposition 2's iteration bound can be negative for an allowed wrong-sign initial coefficient that already satisfies the magnitude condition.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 16 pages, 4 figures