Integration: Thoughts
Roy Schestowitz
Special thanks to Eric Meyer for creating this tool
Overview
- These notes roughly explain how integration fits into AART
- It then explains about a possible framework
- Ultimately, we want integration to be fully-incorporated
The Two Components
- Two stages that the objective function is concerned with:
- Model building
- Model evaluation
- Why do I break it down in this way?
- ...Well, simply because it's easier to comprehend w.r.t. algorithm
Model Building
- This stage takes shape component and intensity component
- It then:
- Calculates a warp (shape) and applies to to intensity
- This involves interpolation
- Some data does not appear in intensity component
Model Evaluation
- Now we need to evaluate the model which was constructed
- If integration was involved, the units here would be different
The Plan and the Questions
- Having chosen a warp, mean shape can be extracted
- From this, mean intensity is obtained
- Can now warp all shapes/intensities to mean
- Integrate each of the couples?
- How to aggregate values from integration?
- How does that form a model
Current Situation
- An argument is passed into all relevant functions
- Several open question
- More theory than practice
- Integration is said to avoid point-hiding
- Hence essential
In Summary
- Too many uncertainties to proceed