English

Mildly-Interacting Fermionic Unitaries are Efficiently Learnable

Quantum Physics 2025-06-18 v2 Data Structures and Algorithms Machine Learning

Abstract

Recent work has shown that one can efficiently learn fermionic Gaussian unitaries, also commonly known as nearest-neighbor matchcircuits or non-interacting fermionic unitaries. However, one could ask a similar question about unitaries that are near Gaussian: for example, unitaries prepared with a small number of non-Gaussian circuit elements. These operators find significance in quantum chemistry and many-body physics, yet no algorithm exists to learn them. We give the first such result by devising an algorithm which makes queries to an nn-mode fermionic unitary UU prepared by at most O(t)O(t) non-Gaussian gates and returns a circuit approximating UU to diamond distance ε\varepsilon in time poly(n,2t,1/ε)\textrm{poly}(n,2^t,1/\varepsilon). This resolves a central open question of Mele and Herasymenko under the strongest distance metric. In fact, our algorithm is much more general: we define a property of unitary Gaussianity known as unitary Gaussian dimension and show that our algorithm can learn nn-mode unitaries of Gaussian dimension at least 2nO(t)2n - O(t) in time poly(n,2t,1/ε)\textrm{poly}(n,2^t,1/\varepsilon). Indeed, this class subsumes unitaries prepared by at most O(t)O(t) non-Gaussian gates but also includes several unitaries that require up to 2O(t)2^{O(t)} non-Gaussian gates to construct. In addition, we give a poly(n,1/ε)\textrm{poly}(n,1/\varepsilon)-time algorithm to distinguish whether an nn-mode unitary is of Gaussian dimension at least kk or ε\varepsilon-far from all such unitaries in Frobenius distance, promised that one is the case. Along the way, we prove structural results about near-Gaussian fermionic unitaries that are likely to be of independent interest.

Keywords

Cite

@article{arxiv.2504.11318,
  title  = {Mildly-Interacting Fermionic Unitaries are Efficiently Learnable},
  author = {Vishnu Iyer},
  journal= {arXiv preprint arXiv:2504.11318},
  year   = {2025}
}

Comments

30 pages, 4 figures

R2 v1 2026-06-28T22:59:18.614Z