Restructure the lecture series: point sources + formalism lectures, chaining into chapter 2, pixelizations as chapter 3, new Scaling Up Lensing chapter 4 - #68
Merged
Conversation
…, pixelizations become chapter 3 Chapter 3 (search chaining) is dissolved: tutorials 1, 2 and 6 move to the end of chapter 2 as tutorials 9 (search chaining), 10 (prior passing) and 11 (SLaM); tutorials 3, 4 and 5 are deleted. Chapter 4 (pixelizations) is renamed chapter 3, with tutorials 5-11 renumbered to 6-12 to make room for a new tutorial 5 (bayesian_formalism, added next). An empty chapter_4_scaling_up_lensing is scaffolded for the new scaling-up chapter. File moves only (plus the no_run.yaml borders rename); content updates to titles, cross-references and prose land in follow-up commits. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
Replaces the outdated numba-centric run-time story with the current reality: JIT compilation of the whole likelihood via JAX (on by default through use_jax=True), CPU vs GPU batching, compilation overhead, and run-time estimation. Chaining is reframed as this chapter's closing tutorials 9-11 and stale HowToGalaxy/chapter references are fixed. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
Replaces the stub with a full lecture: what point sources are, the lens equation as an inverse problem, the triangle-based PointSolver, multiple images and critical curves, magnifications and why fluxes are unreliable, time delays, and how point-source computations differ from the extended-source forward model. Verified against the installed stack and runs end-to-end in both smoke and full-resolution modes. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
…torial tutorial_6_data gains two short sections: one loading and plotting an interferometer dataset (visibilities, uv-plane fitting) that states HowToLens will not cover interferometry further, and one loading and plotting a weak-lensing shear catalogue that defers all detail to chapter 4 as deliberate foreshadowing. Both are backed by new scripts/simulator/ scripts adapted from the workspace simulators, wired in with the standard auto-simulate guard. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
Closes the scaling-up chapter and the core series: what weak lensing is, shear catalogues and their upstream production, the mass-scale ladder from single galaxies to clusters, in-memory simulation of shear data, whisker plots and Kaiser-Squires mass maps, fitting with FitWeak, modeling with AnalysisWeak, tangential shear profiles with a B-mode null test, and joint strong+weak framing. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
Replaces the stub with the series' equations lecture: cosmological distances, the lens equation, convergence and critical surface density, deflection angles, the lensing potential, shear and magnification via the Jacobian, critical curves and caustics, the Einstein radius as the robust mass measurement, and time delays. All lensing mathematics from the rest of the series is deferred here, with every quantity computed and plotted through the live API. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
…orials 9-11 The moved tutorials are retitled and reframed as the closing arc of the lens-modeling chapter, motivated by tutorial 8's run-time discussion. Output paths and search names move to chapter_2; references to the deleted chaining tutorials reroute to the workspace's current chaining and SLaM guides; stale chapter cross-references in tutorials 3-5 now match the new chapter numbering. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
The co-dominant deflector regime: mergers vs line-of-sight alignments, modeling the x2 system with MGE light + isothermal masses, runtime- verified parameter accounting, the fixed-mass-centre trick and the inter-galaxy mass degeneracy, foreshadowing of the shared-halo framing at group scale, and a closing three-lens example backed by a new lens_x3 simulator. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
Opens the scaling-up chapter with the full ladder framing, then teaches the two ways to handle a nearby interloper galaxy: noise scaling its light out of the fit, or modeling it explicitly through the extra_galaxies model interface with centres fixed to the observed light. Central framing is whether the neighbour's light and/or mass affects the main lens analysis. Backed by a new lens_extra_galaxy simulator that also outputs the extra-galaxies mask and centres. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
Replaces per-galaxy mass freedom with a few global relation parameters: Faber-Jackson motivation, luminosity-tied Einstein radii over the x2 dataset, runtime-verified parameter accounting including a composed 12-member model whose dimensionality does not grow, an honest limitations section (scatter, deviant members, anchor anomalies), and the environment-of-single-lenses use case. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
Defines the group scale in LCDM terms (10^13-10^14 Msun halo, BGG, the blurry multi-galaxy/group boundary), introduces tidal truncation and the dPIE mass profile parameter-by-parameter, shows a fit.py-style fit and the effect of adding a group-scale halo, then models members individually and through a scaling relation. Frames group scale as the sweet spot where galaxy-scale tools (including pixelized sources) still work at low dimensionality. Backed by a new group simulator whose members exactly obey the scaling relation. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
…formalism lecture New tutorial_5_bayesian_formalism collects the pixelized-source linear algebra in one place (mapping and blurred mapping matrices, data vector, curvature and regularization matrices, the regularized solve and all five log-evidence terms, computed live against the chapter's dataset). The other twelve tutorials get their new numbers, chapter references matching the restructure, light algebraic framing that defers derivations to tutorial 5, and prose cleanup. tutorial_6_borders' cap-induced failure is fixed (dead sub-border assignments removed) and its NEEDS_FIX skip is dropped from no_run.yaml. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
Defines the cluster regime (30+ member galaxies, >1e14 Msun, many sources at different redshifts), explains multi-plane ray tracing at HowTo level, motivates the move to point-source position fitting, and teaches the CSV interface for composing member catalogues into a factor-graph model. A dedicated results section walks through accessing the multi-plane tracer, per-member profiles, source-plane positions, magnifications and errors. Backed by a new cluster simulator producing the imaging, CSV catalogues and point datasets. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
… structure Chapter README indexes gain tutorials 9-11 (chapter 2), the renumbered pixelizations listing with the new Bayesian formalism tutorial (chapter 3) and a new chapter 4 index; the top-level scripts/notebooks READMEs drop the stale five-chapter/hyper-mode overview in favour of the real four-chapter map; chapter 2's legacy summary file moves past the new tutorials and mentions chaining; chapter 4 gains an introduction file. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
…orrections Adversarial review of the pixelizations chapter: the borders tutorial now loads x2_lens_galaxies at the simulator's real 0.1" pixel scale, actually uses its large-mask demo and matches the simulator's true mass profiles; a stray stale-fit plot is removed from lens modeling; the Bayesian formalism lecture gets the correct second-derivative in the WD03 regularization definition, consistent indices and a Contents entry for its Fit section; adaptive-pixelization pixel counts match the real 28x28 mesh; fit-problems prose stops claiming an Overlay image-mesh; Voronoi->Delaunay in the chapter README and introduction. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
Adversarial review of the parallel-written scaling-up chapter: the tied-member population now consistently uses the scaling_galaxies collection at group scale (matching tutorials 3 and 5, verified against the library's tracer composition); scaling exponents tie back to the Faber-Jackson derivation; source-profile claims match the simulators (ExponentialCoreSph, fixed in lens_x2/lens_x3 headers too); the weak lensing simulator's shear amplitudes now agree with its tutorial's SIS arithmetic; hand-off and teaser sentences across tutorials 1-6 align with what each tutorial actually covers. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
…laims Adversarial review: pre-restructure titles and cross-references fixed (more_ray_tracing was still 'Tutorial 5', the summary 'Tutorial 9'); caustic image-multiplicity statements corrected to the changes-by-two rule with the three-region statement; the fitting tutorial's noise normalization prose now matches its sum-of-logs formula; parameter counts, n_live values and prior/profile names align with the code and simulators (verified by introspection); need-for-speed quotes the real JAX jit log message; HowToGalaxy copy-paste and workspace-path confusion fixed; Contents lists now match every file's actual sections; scores of typos cleaned across both chapters and the optional searches tutorial. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
Notebooks rebuilt from the restructured scripts via PyAutoHands generate.py (new chapter 3/4 trees, updated chapters 1-2, simulator notebooks); llms-full.txt and workspace_index.json regenerated; the curated llms.txt learning path updated to the new chapter map. check_navigator passes (paths + banner lint). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
With the PyAutoHands uncomment fix, generated notebooks now activate 'from autolens import setup_notebook; setup_notebook()' so they run from the workspace root (Colab and local Jupyter alike). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
All nine curated chapter 1 pages rebuilt with their real output images, including the newly written point-sources and lensing-formalism tutorials and the extended data tutorial (interferometer + weak lensing sections). Rendering the dataset-loading tutorials was unblocked by the PyAutoHands setup_notebook uncomment fix. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
This was referenced Aug 6, 2026
CI's Python 3.13 leg deliberately runs the lean no-optional-deps path without jax; the new cluster simulator imports jax for its jitted multi-plane PointSolver, failing that leg (and the cluster tutorial with it via the auto-simulate subprocess). Both now use the established importlib find_spec / sys.exit(0) guard — the same idiom as the two jax-only chapter 3 tutorials — exiting cleanly with an explanatory message when jax is missing, verified both with jax removed (clean skip) and present (full run). Notebooks regenerated. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
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
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
Major restructure of the HowToLens lecture series:
Chapter 1 — Introduction
tutorial_4_point_sourceswritten in full (was a stub): what point sources are, the lens equation as an inverse problem, the triangle-basedPointSolver, multiple images and critical curves, magnifications and why fluxes are unreliable, time delays.tutorial_5_lensing_formalismwritten in full (was a stub): the series' single equations lecture — distances, the lens equation, convergence/Σ_cr, deflections, the potential, shear and the Jacobian, critical curves/caustics, Einstein radius, time delays — all lensing math from other tutorials is deferred here.tutorial_6_datagains one-block interferometer and weak-lensing sections (each backed by a new simulator), stating interferometry is not covered further and deferring weak lensing to chapter 4 as foreshadowing.Chapter 2 — Lens Modeling
tutorial_8_need_for_speedrewritten for the JAX era (JIT compilation,use_jax=Truedefault, GPU, run-time drivers), replacing the outdated numba-centric story.Chapter 3 — Pixelizations (was chapter 4)
tutorial_5_bayesian_formalism: the pixelization equations lecture mirroring the workspace'spixelization/likelihood_function.py(mapping/blurred mapping matrices, data vector D, curvature F, regularization H, the solve, all five evidence terms, WD03/N15/Suyu citations), computed live.no_run.yamlskip dropped.Chapter 4 — Scaling Up Lensing (new)
Six new tutorials: extra galaxies (noise scaling vs explicit modeling), multi-galaxy lenses (x2 modeled with MGE + fixed centres, mass degeneracies, x3 close), scaling relations (luminosity-tied masses, runtime-verified parameter accounting, limitations), group scale (LCDM framing, dPIE truncation, scaling members, the "sweet spot"), cluster scale (multi-plane, point-source rationale, CSV interface, results deep-dive) and weak lensing (shear catalogues, simulate → fit → model, closes the series). Six new simulators support them.
All content was adversarially reviewed (equations, parameter counts and API claims verified against the installed stack by re-execution); READMEs, introductions, the navigator catalogue (
llms.txt/llms-full.txt/workspace_index.json), notebooks and the chapter 1 executed-markdown renders are regenerated.Validation
check_tutorials_complete.pyandcheck_navigator.py: pass.generate.py; markdown renders 9/9 PASS.Merge coordination
CI runs the PyAuto* libraries from same-named branches, so this PR validates against its companions. Suggested merge order: PyAutoGalaxy (mge_model_from centre_fixed fix) and PyAutoHands (setup_notebook uncomment fix) first, then this PR, then the autolens_workspace / PyAutoLens cross-reference PRs. Colab links for new/renamed notebooks are pinned to tag
2026.8.4.1per convention and resolve after the next release'sbump_colab_urls.🤖 Generated with Claude Code
https://claude.ai/code/session_01BxKfSZisjnEn91LRGkN4SU
Generated by Claude Code