2014 Winter Project Week:ParameterSerialization
From NAMIC Wiki
Home < 2014 Winter Project Week:ParameterSerialization
Key Investigators
Matt McCormick, Kitware Steve Pieper, Isomics Jim Miller, GE
Project Description
Objective
- Many analysis algorithms have advanced parameters that must be tuned to work with given datasets.
- Exposing all the advanced parameters in the GUI confuses a user who is only interested in applying the algorithm.
- Exploring an algorithm parameter space is more thorough, efficient, and revealing when done programmically instead of manually in a GUI.
- Provenance with saved parameter files is important for reproducibility.
- The object is to allows CLI modules serialize and deserialize CLI parameters in a JSON file.
- The CLI XML parameters can then be used for "important" or "common" parameters can be tweaked by a user.
- Advanced parameters that deviate from the default can be stored in the file for manual or programmatic adjustment.
- JSON is format that is concise, has good human readability and editability, and widely supported by programming languages.
Approach, Plan
- Create a unit test.
- Add JsonCpp and ParameterSerializer to the Slicer superbuild.
- Enable the SlicerExecutionModel_USE_SERIALIZER in SlicerExecutionModel.
- Consider writing a JSON Schema for CLIs that can be used with a validator.
