English

$O(N^2)$ Universal Antisymmetry in Fermionic Neural Networks

Machine Learning 2022-06-17 v2 Chemical Physics Computational Physics Quantum Physics

Abstract

Fermionic neural network (FermiNet) is a recently proposed wavefunction Ansatz, which is used in variational Monte Carlo (VMC) methods to solve the many-electron Schr\"{o}dinger equation. FermiNet proposes permutation-equivariant architectures, on which a Slater determinant is applied to induce antisymmetry. FermiNet is proved to have universal approximation capability with a single determinant, namely, it suffices to represent any antisymmetric function given sufficient parameters. However, the asymptotic computational bottleneck comes from the Slater determinant, which scales with O(N3)O(N^3) for NN electrons. In this paper, we substitute the Slater determinant with a pairwise antisymmetry construction, which is easy to implement and can reduce the computational cost to O(N2)O(N^2). We formally prove that the pairwise construction built upon permutation-equivariant architectures can universally represent any antisymmetric function. Besides, this universality can be achieved via continuous approximators when we aim to represent ground-state wavefunctions.

Keywords

Cite

@article{arxiv.2205.13205,
  title  = {$O(N^2)$ Universal Antisymmetry in Fermionic Neural Networks},
  author = {Tianyu Pang and Shuicheng Yan and Min Lin},
  journal= {arXiv preprint arXiv:2205.13205},
  year   = {2022}
}

Comments

ICML 2022 AI for Science Workshop