[dependabot]

Updates the requirements on ReducedComplexityModeling to permit the latest version.

Release notes

Sourced from ReducedComplexityModeling's releases.

v0.4.0

ReducedComplexityModeling v0.4.0

Diff since v0.3.1

Breaking Changes

• The data loader moves in from GeometricMachineLearning: the files data_loader.jl, batch.jl, tensor_assign.jl and parametric_data_loader.jl are GML's, with their tests. Batch is exported next to DataLoader; accuracy, number_of_batches, assign_output_estimate and ParametricDataLoader are no longer exported. matrix_assign.jl and mnist_utils.jl are deleted.
• DataLoader(::AbstractMatrix) and DataLoader(::AbstractVector) default to autoencoder = false (a time series); pass autoencoder = true for a set of samples. DataLoader(solution) takes suppress_info as a keyword only.
• DataLoader(tensor, labels::AbstractVector) is removed. Patch the images and build the one-hot target with GMLDatasets, then call DataLoader(input, output), whose output may be an integer array.
• A sample-count mismatch in DataLoader(input, output) throws ArgumentError, not AssertionError.
• accuracy returns float(T) for the loader's input type T, counts one error per wrong label for any integer output type, and divides by every labelled column. accuracy(::NeuralNetwork, ::DataLoader) works.
• ChainRulesCore leaves the dependencies.
• [compat]: AbstractNeuralNetworks 0.9.1, NeuralNetworkParameters 0.5, GeometricBase 0.15, GeometricEquations 0.21.5, GeometricSolutions 0.6.6. Environments on earlier versions, including GeometricBase 0.14, no longer resolve.

Full notes: CHANGELOG.md, section 0.4.0.

Merged pull requests:

• Move the data loader in from GeometricMachineLearning, and drop the fork (#62) (@​michakraus)

Changelog

Sourced from ReducedComplexityModeling's changelog.

[0.4.0] — 2026-10-28

Breaking Changes

• The data loader is this package's now: its code arrives from GeometricMachineLearning. src/data_loader/data_loader.jl, src/data_loader/batch.jl, src/data_loader/tensor_assign.jl and src/data_loader/parametric_data_loader.jl are GeometricMachineLearning's files of those names, moved here, with the testsets that exercise them; GeometricMachineLearning removes its copies of all four. This package's own data_loader.jl, batch.jl, tensor_assign.jl and test/data_loader/draw_batch_for_tensor_test.jl give way to them. src/data_loader/matrix_assign.jl and src/data_loader/mnist_utils.jl are deleted with no successor here: matrix_assign.jl held draw_batch! alone, and no code called it; MNIST data and the patch utilities are GMLDatasets'. test/data_loader/mnist_utils.jl is deleted with them. Batch is now exported next to DataLoader; number_of_batches, assign_output_estimate, convert_input_and_batch_indices_to_array, batch_over_two_axes, accuracy and ParametricDataLoader are unexported, and ParametricDataLoader is public: a caller writes ReducedComplexityModeling.ParametricDataLoader.
• A matrix or a vector is a time series by default. DataLoader(::AbstractMatrix) and DataLoader(::AbstractVector) default autoencoder = false, where the fork defaulted it to true: DataLoader(data) reads the second axis as the number of time steps, and a set of samples is DataLoader(data; autoencoder = true). DataLoader(solution) takes suppress_info as a keyword, so DataLoader(solution, true) raises a MethodError.
• DataLoader(tensor, labels::AbstractVector) is gone. That method preprocessed images for MNIST, with patch_length = 7; patch the images and build the one-hot target with GMLDatasets first, then call DataLoader(input, output), whose output may be an integer array for a tensor, a matrix or a vector input. The fork's pairs took one element type for both arguments, whatever the shape, so an integer output with a real input was a MethodError there.
• A sample-count mismatch in DataLoader(input, output) throws ArgumentError. The third axes of input and output must agree, and the message names both sizes. The fork's @assert threw AssertionError, so a caller that catches that type must catch ArgumentError now.
• accuracy returns float(T), the element type of the loader's input, where the fork returned Float64 whatever the loader held.
• accuracy counts one error per wrong label for an output of any integer type, unsigned ones included, and divides by every labelled column: the output's time steps times its parameter axis. A loader that DataLoader(input, output) builds from a matrix or a vector has one parameter axis entry and many time steps, so its accuracy is a fraction in [0, 1].
• accuracy(::NeuralNetwork, ::DataLoader) works: params is imported from AbstractNeuralNetworks, which closes known issue K12. Known issues K2, K3 and K4 leave KNOWN_ISSUES.md with the code they described.
• ChainRulesCore leaves [deps] and [compat]: the fork's rrule for assign_output_estimate was its only use.
• [compat]: AbstractNeuralNetworks rises to "0.9.1", NeuralNetworkParameters to "0.5", GeometricBase narrows from "0.14.6, 0.15" to "0.15", GeometricEquations rises from "0.21" to "0.21.5", and GeometricSolutions from "0.6" to "0.6.6". The moved code calls networkbackend, params, NamedTupleOfArrays and ArrayOrNamedTuple, which AbstractNeuralNetworks 0.9 defines, so an environment on an earlier AbstractNeuralNetworks no longer resolves this release. AbstractNeuralNetworks 0.9.0 caps NeuralNetworkParameters at 0.4, and 0.9.1 is the first to allow 0.5. NeuralNetworkParameters 0.5 requires GeometricBase 0.15, and GeometricEquations 0.21.5 and GeometricSolutions 0.6.6

... (truncated)

Commits

• 173475c Release 0.4.0
• 67402ac Merge pull request #62 from JuliaRCM/simplify/part-m10
• d00261b Count accuracy errors without unsigned wraparound and over every labelled column
• 2e76209 Reword the CHANGELOG line on the Batch export
• 4900b9c Address the review of PR 62: compat floors, loader argument checks, docstrings
• adfa40f Name this package's manual as the renderer of accuracy's docstrings
• 1c81fff Write train! as the code span it is, and name the doctest's file
• bae05ad Record the loader's user-visible changes, and strip the docstrings' trailing ...
• e5e6768 Cite the device probe's commands in K14 and name batch_over_two_axes
• 41b3590 Move the parametric loader in and bring the loader testsets
• 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)

Summary by CodeRabbit

• Compatibility
  • Expanded compatibility to include version 0.4 of the modeling package, while retaining support for version 0.3.

[coderabbitai]

No actionable comments were generated in the recent review.

Recent review info

Run configuration

• Source mode: Auto
• Repository: Unavailable
• Branch: Unavailable
• File: No configuration file from this repository applied
• Commit: Unavailable
• Layers: defaults
• Review profile: CHILL
• Plan: Advanced
• Run ID: 7de71464-4a03-4b58-aae8-d22ff68ceb37

Commits

Reviewing files that changed from the base of the PR and between 5794199 and 8757144.

Files selected for processing: 1

• Project.toml

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 9 remain after this review.

Walkthrough

The ReducedComplexityModeling compatibility constraint in Project.toml now allows versions 0.3 and 0.4.

Changes: Dependency compatibility

Layer / File(s)	Summary
Widen dependency compatibility Project.toml	The ReducedComplexityModeling compatibility constraint now allows versions 0.3 and 0.4, instead of only 0.3.

Estimated code review effort: 2 (Simple) | ~5 minutes

Change: Other

Possibly related PRs

• JuliaRCM/SparseIdentification.jl#36: Raises the Julia floor to 1.12 and notes compatibility with the Julia 1.11 floor of ReducedComplexityModeling.

Suggested labels: dependencies, julia

Suggested reviewers: michakraus

Merge Risk: Minimal · up to 87571

This only widens the allowed dependency range. Under the current optimizer bounds, ReducedComplexityModeling 0.4 cannot be selected, so environments continue to resolve to 0.3. No behavior change or merge-blocking risk results.

Architecture risk: Low · up to 87571

The change affects 1 system.

Changed systems: Project.toml

Architecture concerns
No architecture-level concerns identified.

Review details

Systems and components

• observed — Project.toml (service) was modified; 1 changed file maps to changed impact.

Before / after behavior

• observed — Modified behavior in Project.toml: The ReducedComplexityModeling compatibility constraint is widened from 0.3 to 0.3, 0.4.

Pre-merge checks: 5

Check name	Status	Explanation
Description Check		Check skipped - CodeRabbit’s high-level summary is enabled.
Title check		The title names the dependency bump for ReducedComplexityModeling and the changed version requirement. The change in Project.toml widens the allowed range from 0.3 to 0.3 and 0.4, which matches the ti…
Docstring Coverage		No eligible functions found in the changed files. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff with cyclomatic complexity of at least 2. Documentat…
Linked Issues check		Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check		Check skipped because no linked issues were found for this pull request.

Full details: Title check

Explanation

The title names the dependency bump for ReducedComplexityModeling and the changed version requirement. The change in Project.toml widens the allowed range from 0.3 to 0.3 and 0.4, which matches the title. The title is specific and describes the main change. It does not mention the breaking changes in 0.4.0, but the title does not need to cover every detail. The wording has a small flaw: the title says 'from 0.3 to 0.3, 0.4', which reads awkwardly, but the meaning is clear.

Full details: Docstring Coverage

Explanation

No eligible functions found in the changed files. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff with cyclomatic complexity of at least 2. Documentation-only edits and deletions within surviving functions are included; unchanged diff context and fully deleted functions are excluded. Analyzed 0 functions across 0 files. (1 skipped: 1 unsupported.)

Finishing Touches

Generate unit tests (beta)

• Commit to this branch
• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.91%. Comparing base (5794199) to head (8757144).

Additional details and impacted files

@@           Coverage Diff           @@
##             main      #37   +/-   ##
=======================================
  Coverage   89.91%   89.91%           
=======================================
  Files          10       10           
  Lines         357      357           
=======================================
  Hits          321      321           
  Misses         36       36           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.