Skip to content

Getting started

MADS Getting Started

Install Julia and MADS (import Pkg; Pkg.add("Mads")) using the installation instruction in the (see also). If you are not familiar with Julia, checkout Julia By Example, learn X in Y minutes, Julia Express). You can also explore Julia examples provided in Mads: examples/learn_julia directory of the Mads.jl repository (github).

To start using MADS, initiate the Julia REPL and execute import Mads to load MADS modules.

All the MADS analyses are based on a MADS problem dictionary that defines the problem.

The MADS problem dictionary is typically loaded from a YAML MADS input file. The loading of a MADS file can be executed as follows:

madsdata = Mads.loadmadsfile("<input_file_name>.mads")

For example, you can execute:

madsdata = Mads.loadmadsfile(Mads.madsdir * "/../examples/getting_started/internal-linear.mads")

The file internal-linear.mads is located in examples/getting_started directory of the Mads.jl repository.

Typically, the MADS problem dictionary includes several classes:

  • Parameters : lists of model parameters
  • Observations : lists of model observations
  • Model : defines a model to predict the model observations using the model parameters

The file internal-linear.mads looks like this:

- a : { init:  1, dist: "Uniform(-10, 10)" }
- b : { init: -1, dist: "Uniform(-10, 10)" }
- o1: { target: -3 }
- o2: { target:  1 }
- o3: { target:  5 }
- o4: { target:  9 }
Model: internal-linear.jl

In this case, there are two parameters, a and b, defining a linear model, f(t) = a * t + b, described in internal-linearmodel.jl.

The Julia file internal-linearmodel.jl is specified under Model in the MADS problem dictionary above.


Mads.showallparameters(madsdata) to show all the parameters.

Mads.showobservations(madsdata) to list all the observations.

MADS can perform various types of analyses:

  • Mads.forward(madsdata) will execute forward model simulation based on the initial parameter values.
  • saresults = Mads.efast(madsdata) will perform eFAST sensitivity analysis of the model parameters against the model observations as defined in the MADS problem dictionary.
  • optparam, iaresults = Mads.calibrate(madsdata) will perform calibration (inverse analysis) of the model parameters to reproduce the model observations as defined in the MADS problem dictionary; in this case, the calibration uses Levenberg-Marquardt optimization.
  • Mads.forward(madsdata, optparam) will perform forward model simulation based on the parameter valuesoptparam` estimated by the inverse analyses above.

More complicated analyses will require additional information to be provided in the MADS problem dictionary. Examples are given in the examples subdirectories of the Mads.jl repository (github).

MADS Command-line execution

MADS can be executed at the command line using madsjl.jl. Link this file in a directory in your search PATH.

For example, using madsjl.jl you can execute:

madsjl.jl diff internal-linear.mads internal-parabola.mads
madsjl.jl internal-parabola.mads forward efast

in the examples/getting_started subdirectory of the Mads.jl repository (github).

MADS Documentation

All the available MADS modules and functions are described at github and readthedocs

Check the files COPYING and LICENSE to see the licensing & copyright information.