Warm-start the imaging potential-correction iterative test#181
Merged
Conversation
subhalo_recovery.py cold-started the iterative LM engine (x0=zeros), which recovered the subhalo (corr 0.78) until PyAutoLens ad97e5fd5 switched LM damping to mu*diag(H) (for the interferometer). Under scale-invariant damping, cold-start from zeros no longer lands in the subhalo basin and the iterative dkappa correlation collapsed to 0.032 (deterministic; failed on main 3.12/3.13, env-independent). Switch to the certified warm-start recipe (PyAutoLens#630): x0 = the one-shot joint solution [s | dpsi], with gauge_project_x0=True so the engine projects the un-gauged one-shot dpsi onto the gauge surface before iterating. The LM now refines inside the one-shot basin (corr 0.7822, evidence 9.07e3, up from cold's 4.21e3) and the gauge constraints hold. Depends on the matching-name PyAutoLens branch (adds gauge_project_x0); library merges first. Verified end-to-end under the smoke env (TEST_MODE=2, DISABLE_JAX=1): all assertions pass. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01V3MxfFWYP7cBPyLGE8MWAV
Jammy2211
pushed a commit
that referenced
this pull request
Jul 17, 2026
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Why
potential_correction/subhalo_recovery.pycold-started the iterative LM engine (x0=zeros), which recovered the subhalo (corr 0.78) until PyAutoLensad97e5fd5switched LM damping tomu*diag(H)(for the interferometer). Under scale-invariant damping, cold-start from zeros no longer lands in the subhalo basin and the iterative dkappa correlation collapsed to a deterministic0.032— failing CI onmain(3.12 & 3.13), env-independent (reproduced with the full smoke env and bare). The one-shot inversion was unaffected (0.78). Bisected toad97e5fd5: the port test was green until the nextiterative.pychange landed.What
Switch to the certified warm-start recipe (PyAutoLens#630):
x0 =the one-shot joint solution[s | dpsi], withgauge_project_x0=Trueso the engine projects the un-gauged one-shot dpsi onto the gauge surface before iterating. The LM now refines inside the one-shot basin —corr 0.7822, evidence9.07e3(up from cold's4.21e3) — and the gauge constraints hold. This is the first test to exercise thex0warm-start feature.Verified end-to-end under the smoke env (
TEST_MODE=2,DISABLE_JAX=1): all assertions pass.Dependency
Requires the matching-name PyAutoLens#632 (adds
gauge_project_x0); the smoke CI clones that branch automatically. Merge the library PR first.🤖 Generated with Claude Code
https://claude.ai/code/session_01V3MxfFWYP7cBPyLGE8MWAV