feat: combined strong+weak lensing example (weak series step 8)

[Jammy2211]

Overview

Step 8 of the weak-lensing series: scripts/weak/features/strong_lensing/{simulator,fit,modeling}.py — weak shear as the large-radius complement to strong lensing of the same mass distribution (Niemiec 2020 hybrid-Lenstool; Oguri 2012 SGAS), the science mode this series targets. Everything it needs is merged: AnalysisWeak (#580), the profile/KS plotters (#582), smoke support (#584).

Plan

• simulator.py: ONE tracer simulated into both an Imaging dataset (central arcs, no lens light) and a WeakDataset (surrounding shear field at larger radii, space-like shape noise) under dataset/weak/strong_lensing/.
• fit.py: load both, fit with a shared tracer via FitImaging + FitWeak, show the joint log likelihood is the sum; visualize both fits + the shear profile.
• modeling.py: joint fit via the standard AnalysisFactor/FactorGraphModel API (graph in NumPy mode — the weak factor is NumPy; imaging analysis JAX internally), Nautilus; prose teaches the complementarity (strong pins the Einstein radius/centre, weak constrains the outer profile — the hybrid-Lenstool insight).
• Auto-simulation via should_simulate; navigator catalogue regenerated in-branch; validation via PYAUTO_TEST_MODE (+ full simulator/fit runs).

Branch: feature/weak-strong-lensing; worktree ~/Code/PyAutoLabs-wt/weak-strong-lensing/ (autolens_workspace only). Autonomy: --auto continuation 2026-07-09, effective supervised — plan-to-issue, park at ship sign-off.

Original Prompt

Click to expand

Combined strong plus weak lensing example. Add autolens_workspace/scripts/weak/features/strong_lensing/ with dedicated simulator.py, fit.py and modeling.py showing weak-lensing shear constraints combined with strong lens modeling of the same mass distribution — the science mode where weak shear is the extra large-radius signal around strong-lens clusters (Niemiec 2020 hybrid-Lenstool, arXiv:2002.04635) and strong-lens groups (Oguri 2012 Sloan Giant Arcs Survey, arXiv:1109.2594), not cosmic shear or galaxy-galaxy lensing. simulator.py simulates the SAME tracer into an imaging dataset (strong) and a surrounding WeakDataset shear field (weak); fit.py fits both with a shared tracer; modeling.py combines AnalysisImaging + AnalysisWeak via PyAutoFit analysis summing to show the joint fit constraining the mass profile better than either alone (parametric core + large-radius shear, the hybrid-Lenstool insight). Depends on AnalysisWeak from feature/weak/4_modeling.md.

[Jammy2211]

Scope extension (discovered mid-implementation): the first mixed-dataset FactorGraphModel (AnalysisImaging + AnalysisWeak) crashed PyAutoFit's combined visualization — the lead factor's VisualizerImaging.visualize_combined receives every factor's analysis and assumed they were all imaging (AttributeError: 'FitWeak' object has no attribute 'data'). Fixed in PyAutoLens (imaging + interferometer visualizers filter to their own analysis type, early-return before plotter construction; regression test added; suite 357 green). This task therefore ships a small library PR alongside the workspace PR, library-first.

[Jammy2211]

Ship sign-off requested (supervised checkpoint — run parked here)

What was done (committed locally, nothing pushed):

• autolens_workspace f-w-s-l branch: scripts/weak/features/strong_lensing/{simulator,fit,modeling}.py + in-branch catalogue regeneration. One tracer → imaging arcs (100×100 @0.1", no lens light) + 400-galaxy shear field to 10" (σ=0.1). modeling.py uses the standard AnalysisFactor/FactorGraphModel API with fully shared priors; prose teaches the hybrid-Lenstool/SGAS complementarity.
• PyAutoLens 7d8139a9 (scope extension, recorded above): mixed-dataset factor graphs crashed combined visualization; imaging + interferometer visualizers now filter to their own analysis type (+regression test).

Validation.

• PyAutoLens full suite: 357 passed.
• simulator.py and fit.py ran end-to-end (imaging LL 6653.7, weak χ² 861/800, joint LL = exact sum 7330.2).
• modeling.py: TEST_MODE mechanics pass, and a full real joint Nautilus fit recovered the truth precisely: einstein_radius = 1.5999 (1.5993–1.6005) vs 1.6, lens ell_comps at truth, source centre/radius/index all within errors. The per-factor max-LH subplots and corner produced.
• Parallel-claim coordination with point-pairing-policies noted in active.md (catalogue regen on whichever merges second).

Heart: YELLOW, same acknowledged reason set as this session's earlier ships.

The question. Approve shipping? On approval: push both branches, open the PyAutoLens PR (pending-release, merge first) and the workspace PR, merge on green, close this issue — then step 7 (Abell 2744 real data, incl. the CSV/FITS WeakDataset loader) starts, the last leg of the home straight.

[Jammy2211]

Shipped ✅ — PyAutoLens#587 (mixed-graph visualization fix) and #251 (the combined strong+weak example) merged library-first, all CI green. The weak series now demonstrates its target science mode end to end: one mass model, arcs inside the Einstein radius + shear to 6 Einstein radii, one joint posterior (er = 1.5999 ± 0.0006 vs truth 1.6). Step 7 (Abell 2744 real data + catalog loaders) is the final leg. Closing.