scipy/conjugate scipy/conjugate

scipy/conjugate `scipy/conjugate`#

Description#

scipy conjugate-gradient minimizer: sparse linear system solver.

This plugin solves the linear system \(A\chi = -\nabla J_0\) that arises in the Gauss-Newton inner loop of a variational inversion, using either:

scipy.sparse.linalg.cg — standard conjugate-gradient (CG).
scipy.sparse.linalg.lobpcg — Locally Optimal Block Preconditioned CG (LOBPCG).

The matrix \(A = \mathbf{I} + \mathbf{H}\mathbf{B}\mathbf{H}^\top\) is never formed explicitly; it is represented as a scipy.sparse.linalg.LinearOperator whose matrix-vector product is evaluated via a TL/forward call to the simulator.

YAML arguments#

The following arguments are used to configure the plugin. pyCIF will return an exception at the initialization if mandatory arguments are not specified, or if any argument does not fit accepted values or type:

Mandatory arguments#

method : str, mandatory: Type of solver.

Optional arguments#

maxiter : int, optional: Maximum number of iterations

maxfun : int, optional: Maximum number of function evaluation

epsg : float, optional: Condition on the norm of the gradient. Converges if ||g||_k < epsg . ||g||_0

Requirements#

The current plugin requires the present plugins to run properly:

Requirement name	Requirement type	Explicit definition	Any valid	Default name	Default version
simulator	Simulator	True	True	gausscost	std

YAML template#

Please find below a template for a YAML configuration:

minimizer:
  plugin:
    name: scipy
    version: conjugate
    type: minimizer

  # Mandatory arguments
  method: XXXXX  # str

  # Optional arguments
  maxiter: XXXXX  # int
  maxfun: XXXXX  # int
  epsg: XXXXX  # float