arXiv:2611.03987
Adaptive Qubit Encoding for Continuous-Variable Quantum Optimization
Samuele Cavinato, Marco Tesoro, Gianluca Morettini, Marta Paiusco, Alessandro Scaggion, Simone Montangero, Ilaria Siloi
Abstract
Optimizing continuous-variable functions on qubit-based quantum hardware requires a finite representation of the search space and a solver-compatible representation of the objective function. Fixed-grid discretizations provide a direct encoding, but restrict the search to a predefined range and resolution; at the same time, constructing a compact Hamiltonian for a general continuous-variable objective can be impractical. Here, we introduce a framework for continuous-variable optimization tailored to qubit-based quantum architectures. The method combines an adaptive discretization with an iterative local mapping of the objective function onto an effective spin-glass Hamiltonian. We assess the approach on unconstrained and constrained instances of the $n$-dimensional Gramacy-Lee benchmark, and on a proof-of-concept inverse-planning problem in intensity-modulated radiotherapy (IMRT) for cancer treatment. The results show that this framework can produce competitive solutions, supporting its use as a flexible mapping strategy for continuous-variable optimization in qubit-based quantum and quantum-inspired pipelines.
AI Summary
The paper addresses how to represent continuous-variable optimization problems on qubit registers when a fixed grid limits range and precision, and a global binary-polynomial objective is difficult to construct. It combines an adaptive discretization with repeatedly fitted local Ising surrogates, allowing objective functions to be supplied through evaluations rather than an explicit polynomial reformulation. Each variable is encoded relative to a moving reference value, with a variable-specific step size adjusted according to optimization history. At each iteration, objective values at neighboring configurations determine the fields and pairwise couplings of a quadratic spin model. A fully connected surrogate on $N=nq$ spins requires $M=N(N+1)/2$ sampled configurations. Candidate solutions are generated using a classical tree-tensor-network (TTN) variational ground-state search and evaluated against the original objective. Numerical tests use separable Gramacy–Lee objectives with $8$–$64$ variables, a sign-penalized variant, and a radiotherapy planning example with $64$ beamlet weights. Most tests use two qubits per variable and TTN bond dimension five. Across ten benchmark runs, TTN improves over best-neighbor selection and simulated annealing on the same representation; comparisons with continuous particle swarm optimization show faster initial progress in several cases. Four qubits per variable do not improve results at equal evaluation budgets. In radiotherapy, particle swarm obtains a lower best objective: $755$ versus $1033$ for TTN. The study demonstrates no quantum-hardware implementation or quantum advantage. Stated limitations include dense surrogate connectivity, surrogate-construction costs, and the reduced, non-clinical radiotherapy example; scalability and reduced-connectivity constructions remain future work.
Error found: The IMRT application claims to enforce non-negative beamlet weights using the constrained benchmark’s penalty strategy, but that benchmark’s explicit +K sign(x) term favors negative weights, and no corrected application-specific formula is supplied. The local Ising interpolation procedure specifies as many sampled configurations as coefficients but does not specify a sampling construction ensuring a full-rank feature matrix or explain how singular systems are handled. The benchmark captions describe logarithmic objective axes although the stated objectives attain negative values, without specifying the transformation needed to plot those values.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 14 pages, 11 figures, comments are welcome