Wavefunction Optimization¶
Wavefunction optimization is a critical step in QMC calculations. It involves adjusting the parameters of the trial wavefunction (Jastrow factors, orbital coefficients, CI coefficients) to minimize the energy or variance.
Overview¶
CHAMP supports several optimization algorithms, controlled by the optwf module.
Key Concepts¶
- Parameters: The variables being optimized (e.g., Jastrow coefficients, orbital rotation parameters).
- Cost Function: The quantity being minimized (usually energy, variance, or a combination).
- Algorithm: The numerical method used to update parameters (e.g., Stochastic Reconfiguration, Linear Method).
Input Keywords¶
The optwf module controls the optimization process.
%module optwf
ioptwf 1 # Enable optimization
ioptjas 1 # Optimize Jastrow parameters
ioptorb 0 # Optimize orbitals (0=no, 1=yes)
ioptci 0 # Optimize CI coefficients
method 'sr_n' # Optimization method (sr_n, lin_d, linear)
nopt_iter 20 # Number of optimization iterations
nblk_max 4000 # Max blocks per iteration
sr_tau 0.05 # Step size for SR
sr_eps 0.001 # Regularization for SR
%endmodule
Available Methods¶
- Energy Minimization: Minimizing the energy using Stochastic Reconfiguration (SR) or other gradient-based methods.
- Linear Method: A robust method that can minimize energy, variance, or a mix.
- Variance Minimization: Minimizing the variance of the local energy (often used for initial optimization).