[Work Log] Work Log - fitting progress
Implemented a image-pyramid-based likelihood and results are much better! Also relaxed the prior, which appears to be overfit.
Run 1:
- parameterization: eigenspace, scaled by sqrt of eigenvalues
- Eigenvectors: 30
- Prior scaling: 36
- Algorithm: pattern-search in two passes termination criteria: mesh size < 0.5 (roughly 1/2 pixel change) First pass: Ran until convergence at pyramid level 3 (i.e. 1/4th original size) Second pass: Ran until convergence at pyramid level 1
After first pass, the global maximum seemed to have been found. The second pass quickly refined it.
This wasn't happening with the prior I trained, which suggests overfitting.
Possible changes to consider
Try using fewer eigenvectors.
Use likelihood mixture rather than product. Should be smoother and allow more non-optimal results.
Consider scaling by inverse of eigenvector dynamic range, so moving by 1 guarantees at least one pixel changing.
Consider using hyperpriors during training.
Train cubic spline covariance model.
Next steps
Extend the track Pass 1: reproject to initialize sampler, run Pass 2: Use reprojected model as prior
Get full dataset results ** run all pairwise on all datasets
Run on detected data