[dependabot]

Updates the requirements on lazymatrix to permit the latest version.

Release notes

Sourced from lazymatrix's releases.

v0.10.0

This version introduces row normalization for dense and sparse matrices, along with ordered normalization recipes. To support both axes consistently, I have reworked the API and taken the opportunity to simplify how matrices store and expose fitted parameters. This brings several breaking changes to the Rust API:

• Normalization, NormalizationParams, and NormalizationStats are renamed ColumnNormalization, ColumnNormalizationParams, and ColumnNormalizationStats. The name NormalizationParams now denotes a trait shared by fitted column, row, and chain parameters.
• LazyMatrix<M, F> and EagerMatrix<M, F> now take a fitted parameter type as their second generic argument. For example, use LazyMatrix<M, ColumnNormalizationParams<F>> for column normalization or LazyMatrix<M, RowNormalizationParams<F>> for row normalization.
• LazyMatrix::from_normalization(data, params) becomes LazyMatrix::from_parts(data, params). The old from_parts(data, centers, scales) becomes from_column_parts(...), with from_row_parts(...) available for rows. The with_centers and with_scales shortcuts are removed.
• Both matrix wrappers expose fitted centers and scales through normalization().centers() and normalization().scales() instead of matrix-level accessors.
• Both wrappers now return (data, params) from into_parts() instead of (data, centers, scales).
• Column and row parameter into_parts() methods now return (dimension, centers, scales), preserving the dimension even when both vectors are absent.
• Submatrix<'a, O> now borrows a logical parent operator instead of storing column normalization separately. Read the parent through as_inner() and selection bounds through row_range() and column_range(). The submatrix's centers() and scales() accessors are removed; inspect the parent's normalization() instead.
• ColumnNormalization and RowNormalization are non-exhaustive. Construct custom recipes with new(center, scale) instead of struct literals.
• Chain fitting now uses public recipe and fitted-parameter bounds, with fit::<M, P>() and fit_apply::<M, P>() replacing scalar generic arguments.

Normalization recipes can now fit parameters separately with .fit() or fit and wrap data with .fit_apply(). Fitted parameters can own their vectors or share them through .view(), so training and validation operators can reuse column normalization without copying parameters or refitting.

Recipes also compose with .then(), with each stage fitting the output of its predecessors. For new samples, .for_new_samples() reuses training column parameters and refits row stages, while .apply_frozen() reapplies all fitted stages unchanged. Chains containing row stages currently require dense storage.

Breaking changes

• stabilize normalization and submatrix APIs (d4a63cd)

... (truncated)

Changelog

Sourced from lazymatrix's changelog.

0.10.0 (2026-11-04)

This version introduces row normalization for dense and sparse matrices, along with ordered normalization recipes. To support both axes consistently, I have reworked the API and taken the opportunity to simplify how matrices store and expose fitted parameters. This brings several breaking changes to the Rust API:

• Normalization, NormalizationParams, and NormalizationStats are renamed ColumnNormalization, ColumnNormalizationParams, and ColumnNormalizationStats. The name NormalizationParams now denotes a trait shared by fitted column, row, and chain parameters.
• LazyMatrix<M, F> and EagerMatrix<M, F> now take a fitted parameter type as their second generic argument. For example, use LazyMatrix<M, ColumnNormalizationParams<F>> for column normalization or LazyMatrix<M, RowNormalizationParams<F>> for row normalization.
• LazyMatrix::from_normalization(data, params) becomes LazyMatrix::from_parts(data, params). The old from_parts(data, centers, scales) becomes from_column_parts(...), with from_row_parts(...) available for rows. The with_centers and with_scales shortcuts are removed.
• Both matrix wrappers expose fitted centers and scales through normalization().centers() and normalization().scales() instead of matrix-level accessors.
• Both wrappers now return (data, params) from into_parts() instead of (data, centers, scales).
• Column and row parameter into_parts() methods now return (dimension, centers, scales), preserving the dimension even when both vectors are absent.
• Submatrix<'a, O> now borrows a logical parent operator instead of storing column normalization separately. Read the parent through as_inner() and selection bounds through row_range() and column_range(). The submatrix's centers() and scales() accessors are removed; inspect the parent's normalization() instead.
• ColumnNormalization and RowNormalization are non-exhaustive. Construct custom recipes with new(center, scale) instead of struct literals.
• Chain fitting now uses public recipe and fitted-parameter bounds, with fit::<M, P>() and fit_apply::<M, P>() replacing scalar generic arguments.

Normalization recipes can now fit parameters separately with .fit() or fit and wrap data with .fit_apply(). Fitted parameters can own their vectors or share them through .view(), so training and validation operators can reuse column normalization without copying parameters or refitting.

Recipes also compose with .then(), with each stage fitting the output of its predecessors. For new samples, .for_new_samples() reuses training column parameters and refits row stages, while .apply_frozen() reapplies all fitted stages unchanged. Chains containing row stages currently require dense storage.

Breaking changes

... (truncated)

Commits

• 2d1c800 chore(release): v0.10.0 (+1 more) (#26)
• 12c1285 docs: scope validation to the affected behavior
• 3c7a273 fix(ci): match the pest psm pin in MSRV checks
• 517be91 perf: accelerate vector and weighted column reductions
• b5f76d0 chore(deps-dev): bump wrangler from 4.152.0 to 4.156.0 in /docs in the websit...
• 52b1bac revert: "chore(release): v0.10.0 (+1 more) (#24)"
• af0b8f8 fix(ci): pin psm for Rust 1.87 checks
• aa5a341 chore(release): v0.10.0 (+1 more) (#24)
• c62adc6 docs: trim the changelog
• 3d784e0 perf: accelerate faer and nalgebra column reductions
• Additional commits viewable in compare view

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

---

Dependabot commands and options

You can trigger Dependabot actions by commenting on this PR:

• @dependabot rebase will rebase this PR
• @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
• @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
• @dependabot ignore this major version will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this minor version will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this dependency will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)

[dependabot]

Superseded by #528.