You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
docs: port Chapter 3 evidence and adaptive reconstruction teaching #101
Bring HowToGalaxy Chapter 3 up to the approved HowToLens teaching standard
Type: docs
Difficulty: large
Autonomy: supervised
Consequence: judge
Target: @howtogalaxy
Original request
can we apply the changes to HowToGalaxy if we did not already happy for you to do this autonomously without me
Scope
Audit current Chapter 3 against HowToLens PR #126. Transfer applicable evidence/Goldilocks explanations, early-content restoration, fit-problem motivation, adaptive reconstruction and bounded-memory experiments. Adapt all science to directly observed galaxies; do not import lens mass degeneracies, borders/ray tracing, Maresca figures or positions penalties. Preserve existing MGE-plus-pixelization model-fit teaching. Regenerate notebooks/catalogues and validate strict profiles, full-size numerical examples, smoke and independent review. Autonomous implementation/shipping authorized; no automatic merge requested.
Plan
Preserve and improve Tutorials 1/2, including complete contents and the explicit log-evidence coefficient-selection rule.
Add Fit Problems followed by Adaptive Reconstruction, and retain the existing combined-model fit as the final lesson with repaired numbering/navigation.
Explain brightness-guided resolution and adaptive smoothing using unlensed galaxy images; keep reconstruction and decomposition caveats explicit.
Validate live API calls, science/numerical examples, generated notebook parity and canonical smoke; run strict environment-profile validation before shipping.
Prepare an open PR after the autonomous shipping gates pass. Tier: judge — merge mode: human /prm.
Implemented and opened PR #102 at 5d98c61 using the ship_workspace workflow. Chapter 3 now includes the evidence/Goldilocks explanation, Fit Problems and Adaptive Reconstruction; the original combined MGE/pixelization model fit is preserved as Tutorial 5.
Validation: all 30 canonical smoke scripts pass; notebook generation/parity/schema/contents, API audit, strict environment declarations, navigation, catalogue determinism and independent review pass. Full-size Tutorials 1/3/4/5 pass. Full-size Tutorial 2 was terminated during its inherited fine-mesh comparison after the revised section completed; its complete capped smoke run passes. Nonlinear searches use test mode 2, so convergence is not assessed.
PR remains open for human review/merge. Heart has the previously acknowledged AutoFit workspace timeout and a stale release-rehearsal gap; no release or CI gate is bypassed.
Merged PR #102 as 2d873b8af0a97f3ac5ddd26f8d09c51ba2153468 after all eight GitHub checks passed, including both Python 3.12 and 3.13 smoke jobs.
Chapter 3 now includes clearer Goldilocks/log-evidence selection, Fit Problems and Adaptive Reconstruction, with the original combined model-fit lesson preserved as Tutorial 5. All 30 local smoke scripts and generated-content checks pass; independent review is CLEAN. The PR documents the optional full-size Tutorial 2 interruption and the test-mode sampler limitation.
Activity
Jammy2211 commentedon Nov 4, 2026
Implemented and opened PR #102 at 5d98c61 using the ship_workspace workflow. Chapter 3 now includes the evidence/Goldilocks explanation, Fit Problems and Adaptive Reconstruction; the original combined MGE/pixelization model fit is preserved as Tutorial 5.
Validation: all 30 canonical smoke scripts pass; notebook generation/parity/schema/contents, API audit, strict environment declarations, navigation, catalogue determinism and independent review pass. Full-size Tutorials 1/3/4/5 pass. Full-size Tutorial 2 was terminated during its inherited fine-mesh comparison after the revised section completed; its complete capped smoke run passes. Nonlinear searches use test mode 2, so convergence is not assessed.
PR remains open for human review/merge. Heart has the previously acknowledged AutoFit workspace timeout and a stale release-rehearsal gap; no release or CI gate is bypassed.
Jammy2211 commentedon Nov 4, 2026
Shipped — 2026-11-04
Merged PR #102 as
2d873b8af0a97f3ac5ddd26f8d09c51ba2153468after all eight GitHub checks passed, including both Python 3.12 and 3.13 smoke jobs.Chapter 3 now includes clearer Goldilocks/log-evidence selection, Fit Problems and Adaptive Reconstruction, with the original combined model-fit lesson preserved as Tutorial 5. All 30 local smoke scripts and generated-content checks pass; independent review is CLEAN. The PR documents the optional full-size Tutorial 2 interruption and the test-mode sampler limitation.