Simple Example

Dakota GUI

First Steps

  1. Make sure the Dakota GUI launches.  At a minimum, you will need Java 1.8 (the JDK version, not the JRE) installed as a prerequisite.
  2. Upon launching the GUI, connect it to Dakota by going to Window > Preferences > Dakota and providing the path to the Dakota executable in the "Dakota executable" field.
  3. Go to File > New Dakota Project to create a new blank Dakota project file.  It is considered good practice to put these project files in a "workspace" directory on your system (such as "Dakota_GUI_Workspace") to stay organized.
  4. Next, we will import one of Dakota's shipped example files into this project.  Make sure your new project is selected in the Dakota Project Navigator pane (the tall pane on the left side of the screen), then go to File > Search Dakota Examples.
  5. Type "rosen" in the search field and click "Search."  Select "rosen_multidim.in" from the list of search results and select the "Add to the currently selected Dakota GUI project" option.  Finally, select "Download."
  6. Once the rosen_multidim.in example is imported, select it in the Dakota Project Navigator pane to change focus to this file in the Settings Editor pane (the tall pane in the center of the screen).
  7. In the Settings Editor pane, click "Run this Dakota study."  You will see Dakota running in the Console pane in the bottom-right corner of the screen.  When Dakota is finished running, a new run result ("Run 1" with a timestamp) will be added to your project.

What should happen

Dakota outputs a large amount of information to help users track progress. Four files should have been created.  These files are available to view by clicking on the run result in the Dakota Project Navigator pane on the left side of the screen.

  1. Dakota Output Stream:  This is Dakota's console output captured to a file.  It contains information about the function evaluations.
  2. Dakota Error Stream:  This is Dakota's error output captured to a file.  It contains information about any errors that may have occurred during the execution of Dakota.  If no errors occurred, this file will be blank.
  3. Tabular Data:  The file rosen_multidim.dat is created due to the specification of tabular graphics data and tabular graphics file. This summarizes the variables and responses for each function evaluation.
  4. Restart File:  The file dakota.rst is a restart file. If a Dakota analysis is interrupted, it can be often be restarted without losing all progress.

What now?

  • For more information on features available in the GUI (such as graphical plotting), consult the Dakota GUI manual.
  • Assuming Dakota ran successfully, skim the three text files (restart files are in a binary format). These are described further in the Dakota User's Manual, Section 2.1.3.
  • This example used a parameter study method, and the Rosenbrock test problem. More details about the example are in Dakota User's Manual, Section 2.3.2 and the test problem is described in Sections 2.3.1 and 21.2.
  • Explore the many methods available in Dakota in the Dakota User's Manual, Chapters 3 - 8.
  • Try running the other examples in the same directory. These are mentioned throughout the Dakota User's Manual and listed in Chapter 2, Table 2.1 for convenience.
  • Learn the syntax needed to use these methods. For running Dakota see the Section 2.4, and for input file information see Section 2.2.
  • Learn how to use your own analysis code with Dakota in Chapter 17.

Dakota on the Command Line

First Steps

  1. Make sure Dakota runs.
    You should see Dakota version information when you type:
    dakota -v
  2. Create a working directory for your Dakota files.
  3. Copy rosen_multidim.in from the $DAKOTA_INSTALL/examples/users directory to the working directory, where $DAKOTA_INSTALL is the Dakota installation directory.
  4. From the working directory, run:
    dakota -i rosen_multidim.in -o rosen_multidim.out rosen_multidim.stdout

What should happen

Dakota outputs a large amount of information to help users track progress. Four files should have been created:

  1. The screen output has been redirected to the file rosen_multidim.stdout. The contents are messages from Dakota and notes about the progress of the iterator (i.e. method/algorithm).
  2. The output file rosen_multidim.out contains information about the function evaluations.
  3. The file rosen_multidim.dat is created due to the specification of tabular graphics data and tabular graphics file. This summarizes the variables and responses for each function evaluation.
  4. The file dakota.rst is a restart file. If a Dakota analysis is interrupted, it can be often be restarted without losing all progress.

In addition to the files, some plots are created due to the specification of graphics. These can be helpful when processing the data or diagnosing unexpected results. Dakota has some data processing capabilities built in. The output file will contain the relevant results. In this case, the output file has details about each of the 81 function evaluations. For more advanced or customized data processing or visualization, the tabular data file can be imported into another analysis tool.

What now?

  • Assuming Dakota ran successfully, skim the three text files (restart files are in a binary format). These are described further in the Dakota User's Manual, Section 2.1.3.
  • This example used a parameter study method, and the Rosenbrock test problem. More details about the example are in Dakota User's Manual, Section 2.3.2 and the test problem is described in Sections 2.3.1 and 21.2.
  • Explore the many methods available in Dakota in the Dakota User's Manual, Chapters 3 - 8.
  • Try running the other examples in the same directory. These are mentioned throughout the Dakota User's Manual and listed in Chapter 2, Table 2.1 for convenience.
  • Learn the syntax needed to use these methods. For running Dakota see the Section 2.4, and for input file information see Section 2.2.
  • Learn how to use your own analysis code with Dakota in Chapter 17.