Repository navigation
feat: rect_adapt_duo — demo of RectangularRotatedAdaptImage multi-source fix - #71
Merged
Merged
Conversation
…lti-source fix
Adds a new demo package under autolens_workspace_developer/rect_adapt_duo/
that visually demonstrates the ghost-peak failure of RectangularSplineAdaptImage
on multi-modal adapt images, and the fix delivered by the new
RectangularRotatedAdaptImage class shipped in PyAutoArray #323.
Scenario: two compact Sersic sources at source-plane (+/-0.5, +/-0.5) behind
an isothermal lens. The diagonal placement is the textbook case where the
separable per-axis CDF produces a 4-quadrant outer-product mesh — two pixel
zones on the real source peaks, two on the empty ghost cross-products.
Run order:
python simulator.py # → dataset/{data,noise_map,psf}.fits
python compare_meshes.py # → output/comparison.png
The headline panels are the source-plane mesh layouts (left column):
baseline shows the textbook 4-quadrant ghost grid; rotated shows a clean
+45° band through both real source peaks. README documents the chi^2 caveat
(rotated case needs different regularization coefficient for like-for-like
fit quality — both meshes need search-tuned regularization for fair
fit-quality comparison, intentionally skipped in this demo for clarity).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
4 tasks done
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
New developer-workspace demo package (
rect_adapt_duo/) that visually demonstrates the multi-source ghost-peak failure inRectangularSplineAdaptImageand the fix delivered by the newRectangularRotatedAdaptImageclass (shipped in PyAutoArray #323).The demo uses two compact Sérsic sources at source-plane
(±0.5, ±0.5)behind an isothermal lens — the textbook case where the separable per-axis CDF produces a 4-quadrant outer-product mesh with two real source zones and two empty ghost cross-products. The new rotated class aligns to the +45° principal axis and concentrates pixels only on the real peaks.Scripts Changed
rect_adapt_duo/__init__.py— package markerrect_adapt_duo/README.md— demo overview, run order, expected result, honest χ² caveatrect_adapt_duo/simulator.py— two-source diagonal lens simulator (isothermal lens, two Sérsic sources at(±0.5, ±0.5), 150×150 grid at 0.05" pixels, Gaussian PSF σ=0.1)rect_adapt_duo/compare_meshes.py— side-by-sideFitImagingfor baseline (RectangularSplineAdaptImage) and new (RectangularRotatedAdaptImage) mesh classes at the true mass model, with truth-derived adapt image. Outputs a 2×3 comparison PNG (source mesh + reconstruction / image-plane model / image-plane residuals, per mesh class)rect_adapt_duo/dataset/{data,noise_map,psf}.fits— generated dataset (small, committed for reproducibility)rect_adapt_duo/output/comparison.png— the comparison figure produced bycompare_meshes.pyUpstream PR
PyAutoLabs/PyAutoArray#323
Test Plan
python simulator.pyproducesdataset/{data,noise_map,psf}.fitspython compare_meshes.pyruns end-to-end and producesoutput/comparison.png(±0.5, ±0.5)zones, two real + two ghost); rotated shows a clean +45° diagonal band concentrating pixels only on the two real source peaksNotes
This is a dev-workspace demo, not a release-pipeline asset. It complements the empirical experiments under
PyAutoArray/files/(ghost_peak_experiment.py,pca_rotation_experiment.py) by exercising the production class through the realFitImagingpipeline end-to-end.🤖 Generated with Claude Code