An identification is found between meta-learning and the problem of determining the ground state of a randomly generated Hamiltonian drawn from a known ensemble. A model-agnostic meta-learning approach is proposed to solve the associated learning problem and a preliminary experimental study of random Max-Cut problems indicates that the resulting Meta Variational Monte Carlo accelerates training and improves convergence.
One-Sentence Summary
Journal version presenting Meta Variational Monte Carlo, which applies meta-learning to accelerate neural quantum state optimization across random Hamiltonian ensembles.
Key Contributions
Meta-learning framework for variational Monte Carlo optimization
Transferable initialization across random Hamiltonian instances
Accelerated training and improved convergence on Max-Cut problems
Cite This Paper If...
You are working on meta-learning for variational Monte Carlo, transferable neural quantum state optimization, or learning optimization strategies across quantum systems. This paper is relevant if you need methods for amortizing the cost of ground-state optimization across families of related Hamiltonians.
Frequently Asked Questions
What is Meta Variational Monte Carlo?
Meta VMC applies model-agnostic meta-learning (MAML) to the variational Monte Carlo framework. Instead of optimizing each Hamiltonian from scratch, the method learns a shared initialization that enables rapid adaptation to new Hamiltonian instances drawn from the same ensemble.
How does this relate to MAML?
The paper identifies a formal connection between the meta-learning objective and the problem of finding ground states for random Hamiltonians. The MAML framework is applied to learn transferable parameter initializations for neural quantum states.
What problems is Meta VMC tested on?
The experimental study focuses on random Max-Cut problems, where the method demonstrates accelerated training convergence compared to standard VMC with random initialization.
How does this differ from the arXiv preprint version?
This is the journal version published in Quantum Machine Intelligence (2023). The arXiv preprint (2011.10614) was posted in 2020. The journal version includes revisions and additional content from the peer review process.
Can Meta VMC be applied beyond Max-Cut?
Yes. The framework is general: any ensemble of Hamiltonians where instances share structural similarities can benefit from meta-learned initializations. This includes quantum chemistry problems, spin models, and other combinatorial optimization tasks.
Keywords
meta variational Monte Carlometa-learning variational Monte Carlotransferable neural quantum stateslearning to optimize VMCmeta-learning quantum many-body systemsneural quantum state transfer
@article{zhao2023meta,
title={Meta-Variational Quantum Monte Carlo},
author={Zhao, Tianchen and Stokes, James and Knitter, Oliver and Chen, Brian and Veerapaneni, Shravan},
journal={Quantum Machine Intelligence},
year={2023}
}
Recommended Citation Contexts
Meta-learning frameworks for quantum Monte Carlo enable rapid convergence on new molecular geometries by leveraging prior optimization experience (Zhao et al., 2023).