jadie1
commented
2 years ago Discussion recap
The first approach for this will be updating the ShapeWorks optimization cost function so that the shape-space entropy (first term) is broken into two parts:
- Entropy for a single time-point across subjects (within time-point entropy)
- Entropy for a single subject across time-points (within sequence entropy)
The configuration-space entropy (second term) will also need to be updated so that instead of summing over just subjects, it sums over subjects and time points.
In order to implement this we will need:
- Support for 4D data in the project spreadsheet and a flag for the optimizer to recognize it
- Support to change the particle update in Python (this will require a way to query subjects and time-points of particles)
Update ShapeWorks optimization so that it can run on 4D data (shapes over time). This will be similar in nature to how we added support for multiple domains.