Image Analyst MKIIWorkflow
HomeWorkflowImage Processing PrinciplesFunctions GlossaryPipelines GlossaryProtocolsQuick How ToSearch

Pipeline Optimizer

The pipeline optimizer allows multiparametric optimization of pipelines using Differential Evolution for an arbitrary goal defined by Excel formula calculations on the pipeline results.
  1. Create or select a pipeline that outputs into Excel Data Window.
  2. Using a Multi-Dimensional Open dialog evaluate one ore more positions. Typical use: evaluate a (statistical) set of positive and a set of negative controls.
  3. In the Excel Data Window open a new worksheet and using formulas enter a calculation (e.g. 1-z factor or p-value) referring to data in the IA Output sheet that will converge to zero with the goal of the optimization.
  4. Erase the contents of the IA Output worksheet and save this Excel file. This will be the template for the optimization.
  5. In the Pipeline Optimizer dialog Optimizer Settings select the template file, and which cell in which worksheet contains the above calculated goal value.
  6. Also set here which positions to evaluate from the multi-dimensional data set.
  7. In the Pipelines Parameters tab checkmark parameters to be optimized and provide a range of optimization with minimum and maximum values.
  8. The optimization is performed on discrete values, so the Steps column adjusts that into how many values is the range divided.
  9. It is advised to save the optimization parameters before start.
  10. Press Optimize. Note: depending on complexity, expect overnight runs as the pipeline will be executed several hundred times.
  11. Use the results list to select the optimum or other sub-optimal parameterings.
  12. Press Store to send the selected optimized set of parameters to the pipeline.
  13. It is advised to open the optimized pipeline using the Pipeline Window button in the main toolbar, and to save it into the My Pipelines folder using the Save As button.