How adamfield118/femmi works
FEMMI maps weak-lensing shear catalogues to convergence estimates: users provide FITS catalogues and survey settings, then receive convergence images, residuals, coverage, and optional noise maps. The central stages are catalogue ingestion, mesh and lensing-operator construction, regularized inverse reconstruction, and survey output. Separate capabilities include neural priors and calibration/comparison workflows. The README documents the survey interface; sampled excerpts show selected implementation details, so some internal connections are omitted where their call paths were not shown.
Main entry points: operators.py Lensing operators · survey.py Survey runner · inverse.py MAP reconstruction · mapping.py Catalogue mapper
Components 21
Survey workflow
- Command interface
femmi/cli.py - Survey runner
femmi/survey.py - FITS ingestion
femmi/io.py - Survey products
femmi/survey_output.py
Reconstruction core
- Catalogue mapper
femmi/mapping.py - Mesh generation
femmi/mesh.py - Finite elements
femmi/elements.py - P3 basis
femmi/basis.py - Element assembly
femmi/assembly.py - Boundary elements
femmi/bem.py - Lensing operators
femmi/operators.py - Observation handling
femmi/observations.py
Priors and inference
- MAP reconstruction
femmi/inverse.py - Regularization
femmi/regularization.py - Prior models
femmi/priors.py - Neural prior
femmi/neural_prior/prior.py - Posterior sampling
femmi/sampling.py
Evaluation tools
- Method calibration
femmi/calibration.py - Paired comparisons
femmi/comparison.py - KS catalogue methods
femmi/catalog.py
Other components
- Survey analyst
Connections 19
- Survey analyst → Command interface: runs mapping
README.md - Survey analyst → Survey runner: uses Python API
README.md - Survey runner → FITS ingestion: reads catalogues
femmi/survey.py - Survey runner → Catalogue mapper: builds mappings
femmi/survey.py - Survey runner → Survey products: writes products
femmi/survey.py - Catalogue mapper → Mesh generation: builds mesh
femmi/mapping.py - Catalogue mapper → Lensing operators: builds operators
femmi/mapping.py - Lensing operators → Mesh generation: uses geometry
femmi/operators.py - Lensing operators → P3 basis: uses basis
femmi/operators.py - Lensing operators → Element assembly: assembles matrices
femmi/operators.py - Lensing operators → Boundary elements: couples boundary
femmi/operators.py - Lensing operators → Finite elements: supports elements
- MAP reconstruction → Observation handling: prepares data
femmi/inverse.py - MAP reconstruction → Regularization: uses penalties
femmi/regularization.py - MAP reconstruction → Prior models: accepts prior
femmi/inverse.py - Prior models → Neural prior: supports learned prior
femmi/priors.py - Posterior sampling → Prior models: uses prior models
femmi/sampling.py - Method calibration → KS catalogue methods: fits KS methods
femmi/calibration.py - Method calibration → Paired comparisons: supports evaluation
femmi/calibration.py