Skip to content

docs: align curvature diagonal floor documentation #444

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Jammy2211 merged 1 commit into from
Aug 13, 2026
Merged

docs: align curvature diagonal floor documentation #444

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
19 changes: 17 additions & 2 deletions autoarray/inversion/inversion/inversion_util.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -46,15 +46,30 @@ def curvature_matrix_with_added_to_diag_from(
It is common for the `curvature_matrix` computed to not be positive-definite, leading for the inversion
via `np.linalg.solve` to fail and raise a `LinAlgError`.

In many circumstances, adding a small numerical value of `1.0e-8` to the diagonal of the `curvature_matrix`
In many circumstances, adding a small numerical value to the diagonal of the `curvature_matrix`
makes it positive definite, such that the inversion is performed without raising an error.

This function adds this numerical value to the diagonal of the curvature matrix.
This function adds the caller-supplied `value` to the diagonal entries selected by
`no_regularization_index_list`. The normal inversion path reads this value from
`Settings.no_regularization_add_to_curvature_diag_value`; the packaged configuration defaults to
`1.0e-3`, and workspaces may override it. The addition is absolute, so its effect depends on the scale of
the curvature matrix.

Parameters
----------
curvature_matrix
The curvature matrix which is being constructed in order to solve a linear system of equations.
value
The numerical value added to each selected diagonal entry.
no_regularization_index_list
The indices of parameters without regularization whose diagonal entries receive `value`.
xp
The array module to use (`numpy` by default; pass `jax.numpy` for JAX support).

Returns
-------
ndarray
The curvature matrix with `value` added to the selected diagonal entries.
"""
if xp.__name__.startswith("jax"):
return curvature_matrix.at[
Expand Down
2 changes: 2 additions & 0 deletions autoarray/settings.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -84,6 +84,8 @@ def __init__(
no_regularization_add_to_curvature_diag_value
If a linear func object does not have a corresponding regularization, this value is added to its
diagonal entries of the curvature regularization matrix to ensure the matrix is positive-definite.
The packaged configuration defaults to `1.0e-3`; workspaces may override it. The addition is absolute,
so its effect depends on the scale of the curvature matrix.
nnls_solver_tol
Convergence tolerance (infinity-norm KKT residual) of the JAX positive-only (NNLS) interior-point
solve. `None` (default) uses jaxnnls's own tolerance ``min(n * eps * 5e3, 1e-2)`` (~1.7e-9 at
Expand Down
Loading