Refactor priors to use explicit GalaxyConfig with catalog-centered distributions #17
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
Replace the hardcoded Euclid priors (PriorConfig with flat sigma values) with the same GalaxyConfig-based YAML schema used by Level 0, making the probabilistic model explicit and configurable for both simple and complex models. Key changes: - Add `center: "catalog"` support to DistributionConfig for data-dependent priors (e.g. flux ~ LogNormal(log(catalog_flux), sigma)) - Add `type: "Offset"` to PositionConfig with dx/dy fields - Create shared `shine/prior_utils.py` with `parse_prior()` function - Replace PriorConfig with `gal: GalaxyConfig` on EuclidInferenceConfig - Rewrite MultiExposureScene._sample_parameters() to be config-driven - Refactor SceneBuilder to use shared parse_prior() - Update configs/euclid_vis.yaml with explicit gal: section https://claude.ai/code/session_01Cfk98zXxwLMsDoGuf5L9Gp
- Document `center: catalog` feature in configuration.md - Document `Offset` position type in configuration.md - Add Euclid VIS config example to configuration.md - Add shine.prior_utils API page and register in mkdocs.yml - Update api/config.md description for new features https://claude.ai/code/session_01Cfk98zXxwLMsDoGuf5L9Gp
Code reviewNo issues found. Checked for bugs and CLAUDE.md compliance. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.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.
Summary
This PR replaces the hard-coded
PriorConfigwith an explicit, flexibleGalaxyConfigthat specifies the full probabilistic model (profile type, parameters, and priors) in configuration. It introduces support for catalog-centered priors via a newcenter="catalog"mechanism inDistributionConfig, allowing per-source catalog values to be used as distribution locations at runtime.Key Changes
New
shine/prior_utils.py: Shared prior-parsing utility that convertsDistributionConfigentries (or fixed numeric values) into NumPyro sample sites. Supports catalog-centered priors by accepting optionalcatalog_valuesparameter.Enhanced
DistributionConfig:center: Optional[str]field to supportcenter="catalog"modecenter="catalog", themeanparameter is optional and ignored; per-source catalog values are used as the distribution location insteadmeanwhencenter="catalog"is set, but still requiresigmaNew
PositionConfig"Offset" mode:dxanddyfields to support position offsets (e.g., from catalog positions)DistributionConfigReplaced
PriorConfigwithGalaxyConfig:shine.euclid.config.PriorConfig(which only held sigma values)EuclidInferenceConfig.priors→EuclidInferenceConfig.gal(aGalaxyConfig)_default_euclid_galaxy_config()factory that builds the default Euclid model matching the previously hard-coded priorsUpdated
MultiExposureScene._sample_parameters():self.config.galinstead ofself.config.priorsparse_prior()to handle both fixed values and distributionscatalog_valuesfor flux and half-light radius (catalog-centered LogNormal priors)Updated
shine.scene.Scene._parse_prior(): Now a static method that delegates toparse_prior()fromprior_utils.Configuration file (
configs/euclid_vis.yaml):priors:section with explicitgal:sectionComprehensive test coverage (
tests/test_prior_utils.py):Notable Implementation Details
center="catalog", the median is set to each source's catalog value; internallylog(catalog_value)is used as the log-space location parameter.GalaxyConfigfactory reproduces the exact priors that were previously hard-coded, ensuring existing inference results are reproducible.https://claude.ai/code/session_01Cfk98zXxwLMsDoGuf5L9Gp