Linear Method¶

The Linear Method is a powerful optimization algorithm that expands the wavefunction to first order in the parameter changes and solves the resulting generalized eigenvalue problem.

Overview¶

The Linear Method can be more stable and efficient than simple gradient descent, especially for optimizing many parameters simultaneously (e.g., Jastrow + Orbitals).

It is selected with method = 'linear' or method = 'lin_d'.

Key Features¶

• Robustness: Can handle large parameter sets.
• Efficiency: Often requires fewer iterations than simple gradient methods.
• Cost Function: Can target energy, variance, or a mix.

Usage Example¶

%module optwf
    ioptwf      1
    method      'linear'
    nopt_iter   10
%endmodule