Answers are grounded in this paper and cite their sources.
Summary. A comparative, model-level study of four quantum algorithms ported to a common photonic continuous-variable (CV) model: control-free quantum phase estimation (QPE), randomized product formulas (qDRIFT), quantum Kolmogorov-Arnold networks (QKAN), and Regev's factoring algorithm. Algorithms are compared under a shared noise channel combining exact Fock-space photon loss and Gaussian displacement, with Gottesman-Kitaev-Preskill (GKP) error correction. Main findings. Control-free QPE is the most noise-robust because it needs no controlled unitaries and, for quadratic Hamiltonians, no non-Gaussian gates; the advantage survives for non-Gaussian Hamiltonians. GKP correction helps only once the physical noise exceeds the finite-squeezing floor (proved), and whether an algorithm benefits is set by a noise- versus systematic-limited dichotomy. The core GKP correction model is validated against an explicit Glancy-Knill error-correction Monte Carlo (agreement ≤0.1%). A discrete-variable surface-code baseline shows the CV advantage is structural (Gaussian operations are free), not threshold-based. Scope. This is a model-level numerical comparison, not an experiment; the GKP correction is a validated surrogate rather than a full Fock-space error-correction circuit, and instances are modest in size. Results, sensitivity analyses, and a two-dimensional noise phase diagram are included. Reproducibility. Standalone scripts and released JSON results reproduce every figure and table.
No takes yet. Share an insight, caveat, or question.
Wakaura et al. (2026) studied this question.