Improved Optimization for the Neural-network Quantum States and Tests on the Chromium Dimer
arXiv:2404.09280 · doi:10.1063/5.0214150
Abstract
The advent of Neural-network Quantum States (NQS) has significantly advanced wave function ansatz research, sparking a resurgence in orbital space variational Monte Carlo (VMC) exploration. This work introduces three algorithmic enhancements to reduce computational demands of VMC optimization using NQS: an adaptive learning rate algorithm, constrained optimization, and block optimization. We evaluate the refined algorithm on complex multireference bond stretches of and within the cc-pVDZ basis set and calculate the ground-state energy of the strongly correlated chromium dimer () in the Ahlrichs SV basis set. Our results achieve superior accuracy compared to coupled cluster theory at a relatively modest CPU cost. This work demonstrates how to enhance optimization efficiency and robustness using these strategies, opening a new path to optimize large-scale Restricted Boltzmann Machine (RBM)-based NQS more effectively and marking a substantial advancement in NQS's practical quantum chemistry applications.
13 pages, 9 figures, and 2 tables
References in corpus (43)
- Adam: A Method for Stochastic Optimization
- The density-matrix renormalization group in the age of matrix product states
- Solving the Quantum Many-Body Problem with Artificial Neural Networks
- Recent developments in the PySCF program package
- Heat-bath Configuration Interaction: An efficient selected CI algorithm inspired by heat-bath sampling
- Deep neural network solution of the electronic Schrödinger equation
- Ab-Initio Solution of the Many-Electron Schrödinger Equation with Deep Neural Networks
- Semistochastic Heat-bath Configuration Interaction method: selected configuration interaction with semistochastic perturbation theory
- Generalized Lanczos Algorithm for Variational Quantum Monte Carlo
- The Density Matrix Renormalization Group in Chemistry and Molecular Physics: Recent Developments and New Challenges
- A deterministic alternative to the full configuration interaction quantum Monte Carlo method
- Optimization of quantum Monte Carlo wave functions by energy minimization
- Symmetries and many-body excited states with neural-network quantum states
- Fermionic neural-network states for ab-initio electronic structure
- Weak binding between two aromatic rings: feeling the van der Waals attraction by quantum Monte Carlo methods
- Cheap and near exact CASSCF with large active spaces
- Modern Approaches to Exact Diagonalization and Selected Configuration Interaction with the Adaptive Sampling CI Method
- Semistochastic Projector Monte Carlo Method
- CASSCF with Extremely Large Active Spaces using the Adaptive Sampling Configuration Interaction Method
- Quantum Machine Learning for Electronic Structure Calculations
- Fast Semistochastic Heat-Bath Configuration Interaction
- Ab-initio quantum chemistry with neural-network wavefunctions
- Linear-scaling and parallelizable algorithms for stochastic quantum chemistry
- Variational Monte Carlo calculations of nuclei with an artificial neural-network correlator ansatz
- Unbiasing the initiator approximation in Full Configuration Interaction Quantum Monte Carlo
- Precise measurement of quantum observables with neural-network estimators
- Towards the ground state of molecules via diffusion Monte Carlo on neural networks
- Neural Error Mitigation of Near-Term Quantum Simulations
- An explicitly correlated approach to basis set incompleteness in Full Configuration Interaction Quantum Monte Carlo
- Orbital Optimization in Selected Configuration Interaction Methods
- Machine learning configuration interaction for ab initio potential energy curves
- Solving Quasiparticle Band Spectra of Real Solids using Neural-Network Quantum States
- Accurate many-body electronic structure near the basis set limit: application to the chromium dimer
- Deep Learning-Enhanced Variational Monte Carlo Method for Quantum Many-Body Physics
- A Projector Quantum Monte Carlo Method for non-linear wavefunctions
- A Self-Attention Ansatz for Ab-initio Quantum Chemistry
- Complementary First and Second Derivative Methods for Ansatz Optimization in Variational Monte Carlo
- Symmetry projected Jastrow mean field wavefunction in variational Monte Carlo
- Neural network backflow for ab-initio quantum chemistry
- Ground state search by local and sequential updates of neural network quantum states
- A Non-stochastic Optimization Algorithm for Neural-network Quantum States
- Reduced Scaling Hilbert Space Variational Monte Carlo
- Stochastic perturbation theory to correct non-linearly parametrized wavefunctions