Repository navigation
Conversation
The unification onto _BaseScorePeriodicMLP gave ScorePeriodicMLP the same odd=False Fourier config as ScoreSymmetricPeriodicMLP, making the two architectures identical and contradicting ScorePeriodicMLP's own sine+cosine docstring. Restore odd=True and add regression tests that fail without the fix.
Covers the new even/odd/idx_offset parameters on get_fourier_func and the new 2D expansion get_fourier_func2d, neither of which had any tests. Reconstruction accuracy is checked against known analytic periodic functions.
Covers the two previously-untested N-dimensional generalizations: the v_sigma finite-difference Fokker-Planck regularization path, and the dim==2 branch of the end-of-training free-energy-landscape logging.
Demonstrates building a score network invariant under a custom central-inversion symmetry (odd*odd and even*even products of per-coordinate Fourier bases) and plugging it into FPSL via score_model, the same extension pattern used for the custom noise schedule earlier in the notebook.
This branch has not been deployed
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Port of the N-dim generalization work from the private fork:
v_sigma-basedFP regularization (finite-difference estimator, replaces the 1D-only exact
gradient when set), N-dim W&B FEL logging,
get_fourier_func2d, and theget_fourier_func/FourierFeatureseven/odd cleanup. Left outCustomCentralInversionMLP/FPSL2Don purpose — turned that into a tutorialsection instead of new API surface, since nothing used it yet.
Also: found and fixed a real bug from the private-side refactor —
ScorePeriodicMLPhad silently become cosine-only (same asScoreSymmetricPeriodicMLP), losing its sine terms. Added tests that catchit.
Added test coverage for everything above (none of it had tests before) plus
a changelog entry.