Learning Minimal Representations of Fermionic Ground States
Preprint
Abstract
We introduce an unsupervised machine-learning framework that discovers optimally compressed representations of quantum many-body ground states. Using an autoencoder neural network architecture on data from $L$-site Fermi-Hubbard models, we identify minimal latent spaces with a sharp reconstruction quality threshold at $L-1$ latent dimensions, matching the system's intrinsic degrees of freedom. We demonstrate the use of the trained decoder as a differentiable variational ansatz to minimize energy directly within the latent space. Crucially, this approach circumvents the $N$-representability problem, as the learned manifold implicitly restricts the optimization to physically valid quantum states.
Keywords
All Related Versions
- Version 1, 2025-12-12, ArXiv
- Published version: Physical Review Research, 8 (3).
This publication has 53 references indexed in Scilit:
- Parallel Implementation of the Density Matrix Renormalization Group Method Achieving a Quarter petaFLOPS Performance on a Single DGX-H100 GPU NodeJournal of Chemical Theory and Computation, 2024
- Massively parallel quantum chemical density matrix renormalization group methodJournal of Computational Chemistry, 2020
- Approaching the Basis Set Limit of CCSD(T) Energies for Large Molecules with Local Natural Orbital Coupled-Cluster MethodsJournal of Chemical Theory and Computation, 2019
- PySCF: the Python‐based simulations of chemistry frameworkWIREs Computational Molecular Science, 2017
- DFT in a nutshellInternational Journal of Quantum Chemistry, 2012
- The density-matrix renormalization group in the age of matrix product statesAnnals of Physics, 2011
- Molecular Electronic‐Structure Theory2000
- An Introduction to Coupled Cluster Theory for Computational Chemists2000
- On the limited memory BFGS method for large scale optimizationMathematical Programming, 1989
- Simulating physics with computersInternational Journal of Theoretical Physics, 1982