Echo-QAOA: Constructive Spin-Echo Response Design and Per-Qubit Marginal Distillation for Variational Quantum Optimization
Abstract Solving constrained combinatorial optimization problems such as Maximum Weight Independent Set (MWIS) via the Quantum Approximate Optimization Algorithm (QAOA) requires navigating rugged 2p-dimensional landscapes shaped by competing couplings J and heterogeneous local fields h. Homogeneous annealing schedules lack site-specific phase resolution and require costly online circuit selection, whereas direct neural angle regression collapses under cross-branch jumping and an implicit-function curvature conflict between peak flatness and coordinate stability. We propose Echo-QAOA, exploiting global spin-flip parity conjugation to cancel quadratic couplings H_J exactly within the unmodified QAOA ansatz, factorizing target probabilities into O(kn) single-qubit responses and enabling a zero-measurement overlap decoder that transforms ill-conditioned angle regression into permutation-equivariant marginal distillation (Echo-Prior). Using Daubechies spectral factorization, we prove that arbitrary-order Flat-Top Echo responses are realizable with real angles, achieve O(delta^(2k)) peak error tolerance, and remain strictly convex on zero-free intervals. Across 2,750 five-family benchmark and 5,120 cross-scale MWIS instances, Echo-Prior-Flat outperforms annealing initialization and 36 angle-regression models, expanding P_gs >= 50% coverage by 4.9x with zero Stage-1 quantum measurements.
Authors
- Jianxuan Yao (ORCID: https://orcid.org/0009-0007-6283-7905)
Institutions
- ShanghaiTech University (CN)
Publication Details
- Journal
- Research Square
- Published
- 2026-11-13
- DOI
- https://doi.org/10.21203/rs.3.rs-11707296/v1
- Primary Topic
- Quantum Computing Algorithms and Architecture
- Type
- preprint