Feature/matern adaptive by Jammy2211 · Pull Request #214 · Jammy2211/PyAutoArray

Jammy2211 · 2026-02-13T11:30:20Z

This pull request introduces a new adaptive regularization scheme based on a Matérn kernel that adapts to pixel brightness, along with supporting infrastructure for efficient computation and testing. The main focus is to enable more flexible, data-driven regularization in pixelized inversions, particularly for use cases that benefit from spatially adaptive smoothing. Additionally, several utility and interface improvements were made to support the new regularization and enhance backend compatibility.

New adaptive Matérn kernel regularization:

Added MaternAdaptiveBrightnessKernel class in matern_adaptive_brightness_kernel.py, which implements a Matérn kernel regularization scheme where per-pixel weights adapt to the reconstructed brightness, enabling spatially varying smoothing. Includes methods for constructing the covariance matrix and computing weights, as well as detailed documentation.
Registered the new regularization class in the module's __init__.py for import.
Added comprehensive unit tests for the new regularization in test_matern_adaptive_brightness_kernel.py.

Backend and kernel improvements:

Refactored the Matérn kernel implementation to use a backend-agnostic gamma function (gamma_xp), ensuring compatibility with both NumPy and JAX. [1] [2]
Relaxed type casting in MaternKernel to allow for non-float parameters, improving flexibility and backend support.

Delaunay triangulation enhancements for Matérn regularization:

Added minimal Delaunay triangulation functions (scipy_delaunay_matern, jax_delaunay_matern) optimized for Matérn regularization, returning only the necessary information for downstream computation.
Updated the Delaunay interface logic to optionally skip Voronoi area calculations for cases where only minimal triangulation is needed, controlled by a new skip_areas flag. [1] [2]
Added the skip_areas flag to the Preloads class to allow this optimization to be configured. [1] [2]

Copilot

Pull request overview

This pull request introduces a new adaptive regularization scheme (MaternAdaptiveBrightnessKernel) that combines Matérn kernel-based regularization with brightness-adaptive weights, enabling spatially varying smoothness in pixelized inversions. The implementation is designed to work with both NumPy and JAX backends and includes optimizations for Delaunay triangulation when area calculations are not needed.

Changes:

Added MaternAdaptiveBrightnessKernel class with brightness-adaptive weighting for Matérn regularization
Introduced backend-agnostic gamma_xp function and relaxed type constraints in MaternKernel for improved backend compatibility
Added minimal Delaunay triangulation functions and skip_areas flag to optimize performance when Voronoi areas are not required

Reviewed changes

Copilot reviewed 6 out of 7 changed files in this pull request and generated 10 comments.

Show a summary per file

File	Description
`autoarray/inversion/regularization/matern_adaptive_brightness_kernel.py`	New adaptive Matérn regularization class with brightness-dependent weights and weighted covariance matrix construction
`autoarray/inversion/regularization/__init__.py`	Registered the new `MaternAdaptiveBrightnessKernel` class for import
`autoarray/inversion/regularization/matern_kernel.py`	Added backend-agnostic `gamma_xp` function and relaxed type casting for scale/nu parameters
`autoarray/structures/mesh/delaunay_2d.py`	Added minimal Delaunay functions for Matérn usage and implemented `skip_areas` flag logic
`autoarray/preloads.py`	Added `skip_areas` parameter to control Delaunay area computation
`autoarray/structures/arrays/kernel_2d.py`	Fixed parameter annotation spacing
`test_autoarray/inversion/regularizations/test_matern_adaptive_brightness_kernel.py`	Comprehensive tests for the new regularization scheme

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

autoarray/inversion/regularization/matern_adaptive_brightness_kernel.py

Copilot · 2026-02-13T11:36:51Z

autoarray/preloads.py

        linear_light_profile_blurred_mapping_matrix=None,
        use_voronoi_areas: bool = True,
        areas_factor: float = 0.5,
+        skip_areas: bool = False,


The newly added skip_areas parameter is not documented in the Parameters section of the Preloads docstring. Please add documentation explaining what this parameter controls, when it should be set to True, and how it affects the Delaunay triangulation computation (specifically that it skips Voronoi area calculations and split point computations).

@copilot open a new pull request to apply changes based on this feedback

test_autoarray/inversion/regularizations/test_matern_adaptive_brightness_kernel.py

Copilot · 2026-02-13T11:36:52Z

autoarray/inversion/regularization/matern_adaptive_brightness_kernel.py

+        coefficient
+            The regularization coefficient which controls the degree of smooth of the inversion reconstruction.
+        scale
+            The typical scale of the exponential regularization pattern.


The parameter description refers to "exponential regularization pattern" but this class implements a Matérn kernel regularization. The description should refer to "Matérn regularization pattern" for accuracy. While the parent class has the same issue, it should be corrected in this new class.

Suggested change

The typical scale of the exponential regularization pattern.

The typical scale of the Matérn regularization pattern.

autoarray/inversion/regularization/matern_adaptive_brightness_kernel.py

autoarray/structures/mesh/delaunay_2d.py

Copilot · 2026-02-13T11:36:53Z

test_autoarray/inversion/regularizations/test_matern_adaptive_brightness_kernel.py

+    mesh_grid = aa.m.MockMeshGrid(neighbors=neighbors, neighbors_sizes=neighbors_sizes)
+


Variable mesh_grid is not used.

Suggested change

mesh_grid = aa.m.MockMeshGrid(neighbors=neighbors, neighbors_sizes=neighbors_sizes)

…kernel.py Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Copilot · 2026-02-13T12:56:51Z

@Jammy2211 I've opened a new pull request, #215, to work on those changes. Once the pull request is ready, I'll request review from you.

…brightness_kernel.py Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

… parameters Co-authored-by: Jammy2211 <23455639+Jammy2211@users.noreply.github.com>

Document skip_areas parameter in Preloads docstring

Copilot

Pull request overview

Copilot reviewed 6 out of 7 changed files in this pull request and generated no new comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

…kernel.py Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Jammy2211 added 3 commits February 13, 2026 10:47

added adaptive matern kernel

09be29c

unit tests added with updates for JAX and correct functionality

ac9805b

end to end matern tests and skip areas added

e366b49

Jammy2211 requested a review from Copilot February 13, 2026 11:30

Copilot started reviewing on behalf of Jammy2211 February 13, 2026 11:30 View session

Copilot AI reviewed Feb 13, 2026

View reviewed changes

Jammy2211 and others added 4 commits February 13, 2026 12:56

Update autoarray/inversion/regularization/matern_adaptive_brightness_…

f724404

…kernel.py Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Update autoarray/inversion/regularization/matern_adaptive_brightness_…

9184f28

…kernel.py Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Update autoarray/inversion/regularization/matern_adaptive_brightness_…

a14f39a

…kernel.py Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Initial plan

244caf0

Copilot AI mentioned this pull request Feb 13, 2026

Document skip_areas parameter in Preloads docstring #215

Merged

Jammy2211 and others added 3 commits February 13, 2026 12:56

Update test_autoarray/inversion/regularizations/test_matern_adaptive_…

12de13b

…brightness_kernel.py Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Add documentation for skip_areas, use_voronoi_areas, and areas_factor…

29a2d89

… parameters Co-authored-by: Jammy2211 <23455639+Jammy2211@users.noreply.github.com>

Merge pull request #215 from Jammy2211/copilot/sub-pr-214

e102dc7

Document skip_areas parameter in Preloads docstring

Jammy2211 requested a review from Copilot February 13, 2026 13:26

Copilot started reviewing on behalf of Jammy2211 February 13, 2026 13:26 View session

Copilot AI reviewed Feb 13, 2026

View reviewed changes

Jammy2211 and others added 2 commits February 13, 2026 13:30

Update autoarray/inversion/regularization/matern_adaptive_brightness_…

917b78c

…kernel.py Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Update autoarray/structures/mesh/delaunay_2d.py

95cf5b7

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Jammy2211 merged commit b2dd830 into main Feb 13, 2026
8 checks passed

Jammy2211 deleted the feature/matern_adaptive branch February 13, 2026 13:39

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Feature/matern adaptive#214

Feature/matern adaptive#214
Jammy2211 merged 12 commits intomainfrom
feature/matern_adaptive

Jammy2211 commented Feb 13, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI Feb 13, 2026

Uh oh!

Jammy2211 Feb 13, 2026

Uh oh!

Uh oh!

Copilot AI Feb 13, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI Feb 13, 2026

Uh oh!

Copilot AI commented Feb 13, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

	The typical scale of the exponential regularization pattern.
	The typical scale of the Matérn regularization pattern.

		mesh_grid = aa.m.MockMeshGrid(neighbors=neighbors, neighbors_sizes=neighbors_sizes)

Conversation

Jammy2211 commented Feb 13, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI Feb 13, 2026

Choose a reason for hiding this comment

Uh oh!

Jammy2211 Feb 13, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Copilot AI Feb 13, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI Feb 13, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI commented Feb 13, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants