Optimize correspondence function, add fixed domain precomputation #2483
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
When optimizing with fixed domains (e.g., adding new shapes to an existing model), precompute the shape space from fixed shapes to accelerate gradient computation. This avoids recomputing SVD and covariance matrices each iteration when most shapes are fixed. Also simplifies correspondence functions by removing unused inverse covariance matrix computation - the gradient is computed directly as Y × pinvMat per the thesis derivation. Additionally, there was a bug in the CorrespondenceFunction (but not Legacy one) where the sample block was used for all energy computations, sot he per-particle energy values were incorrect.
The LegacyCorrespondenceFunction is used when normals are off.
Extends fixed shape space precomputation to support both
correspondence function types (mesh-based and legacy ensemble
entropy). When optimizing with mostly fixed domains (e.g., adding 1
new shape to 200 existing), precomputes the covariance structure from
fixed shapes to avoid expensive SVD recomputation each iteration.
- Add PrecomputeForFixedDomains() and Schur complement gradient to
LegacyCorrespondenceFunction for legacy/ensemble entropy mode
- Implement lazy precomputation that triggers on first iteration when
shape matrices are populated
- Select correct function based on mesh_based_attributes setting
- Add distinct profiling scopes (FixedShapeSpace, MeanEnergy, Entropy)
to differentiate code paths
When optimizing with fixed domains (e.g., adding new shapes to an existing model), precompute the shape space from fixed shapes to accelerate gradient computation. This avoids recomputing SVD and covariance matrices each iteration when most shapes are fixed. Also simplifies correspondence functions by removing unused inverse covariance matrix computation - the gradient is computed directly as Y × pinvMat per the thesis derivation. Additionally, there was a bug in the CorrespondenceFunction (but not Legacy one) where the sample block was used for all energy computations, sot he per-particle energy values were incorrect.
The LegacyCorrespondenceFunction is used when normals are off.
Extends fixed shape space precomputation to support both
correspondence function types (mesh-based and legacy ensemble
entropy). When optimizing with mostly fixed domains (e.g., adding 1
new shape to 200 existing), precomputes the covariance structure from
fixed shapes to avoid expensive SVD recomputation each iteration.
- Add PrecomputeForFixedDomains() and Schur complement gradient to
LegacyCorrespondenceFunction for legacy/ensemble entropy mode
- Implement lazy precomputation that triggers on first iteration when
shape matrices are populated
- Select correct function based on mesh_based_attributes setting
- Add distinct profiling scopes (FixedShapeSpace, MeanEnergy, Entropy)
to differentiate code paths
…s into fixed_shape_space
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fixed domains should construct covariance matrix from only fixed domains #2179
Speed up correspondence function #2482
per thesis equation 2.35