Variational inversions

Variational inversion

Description

This mode computes a full variational inversion based on measurements, a corresponding observation vector, and a control vector. This mode needs a minimizer to be computed, which itself requires a so-called ‘’simulator’’, i.e. the function to minimize (at the moment, the default function is the Bayesian Gaussian cost function).

Details on the minimizers are given here.

The Yaml paragraph corresponding to a variational run is as follows:

mode:
  plugin:
    name: 4dvar
    version: std
  minimizer :
    plugin:
      name: congrad
      version: std
    simulator:
      plugin:
        name: gausscost
        version: std
    maxiter: 10
    epsg: 0.03
    df1: 0.01
    kverbose: 2

Requirements

A variational inversion requires the following plugins to be executed properly:

1. a control vector to define the control vector shape and corresponding operations; mandatory

2. an observation vector to define observations to be compared with; mandatory

3. an minimizer to optimize the function; optional; default is (m1qn3, std)

The following plugins are indirectly needed to compute a variational inversion:

1. a numerical model; mandatory

2. an observation operator to compute \(\mathbf{x} \rightarrow \mathcal{H}(\mathbf{x})\) and its adjoint; optional: default is (standard, std)

Example