AdaptiveResonance.jl

AdaptiveResonance v0.8.4

Diff since v0.8.3

This patch release changes the usage and conventions of internal docstring variables so as to quiet warnings from previous versions and in certain circumstances.

Merged pull requests:

• 132 fix variant statement warning (#133) (@AP6YC)
• Release/v0.8.4 (#134) (@AP6YC)

Closed issues:

• Fix variant statement warning (#132)

AdaptiveResonance v0.8.3

Diff since v0.8.2

This patch makes minor changes throughout both the hosted documentation and code.

Merged pull requests:

• CompatHelper: add new compat entry for Statistics at version 1, (keep existing compat) (#127) (@github-actions[bot])
• Feature/mergeart (#128) (@AP6YC)
• Version cleanup (#130) (@AP6YC)
• Release/v0.8.3 (#131) (@AP6YC)

Closed issues:

• MergeART (#122)
• Version cleanup (#129)

AdaptiveResonance v0.8.2

Diff since v0.8.1

Closed issues:

• Add downloaded assets to deployed docs folder (#123)

Merged pull requests:

• Docs assets (#124) (@AP6YC)
• Release/v0.8.2 (#125) (@AP6YC)

AdaptiveResonance v0.8.1

Diff since v0.8.0

Closed issues:

• Readme updates (#113)
• Update tagbot to correct release branch (#114)
• Add github issue templates (#115)
• Move binaries to external project (#119)

Merged pull requests:

• Feature/readme updates (#116) (@AP6YC)
• Feature/GitHub issue templates (#117) (@AP6YC)
• Specify master branch for tagbot (#118) (@AP6YC)
• Feature/remove git lfs (#120) (@AP6YC)
• Release/v0.8.1 (#121) (@AP6YC)

AdaptiveResonance v0.8.0

Release notes:

This minor release overhauls a lot of the internals of the package while preserving the main API of the package. This release includes:

• Activation, match, and update functions can be swapped out within modules freely using Julia's metaprogramming techniques without any overhead.
• Runtime and memory footprints for most modules are significantly reduced with various optimizations after in-depth profiling.
• Several ART modules are refactored such that other modules are actually variants of base classes (i.e., Gamma-Normalized FuzzyART is a FuzzyART module with activation=:gamma_activation and match=:gamma_match, Default ARTMAP is an ARTMAP with activation=:choice_by_difference, etc.).
• Statistics are logged in a common stats field of ART modules, which contain details of interest such as the most recent best-matching unit index, activation, and match values.
• PrecompileSignatures.jl is added to the project to automatically compile function signatures with concrete types at install time to slightly increase time-to-first-run during runtime.

Diff since v0.7.0

Closed issues:

• Add PrecompileSignatures (#105)
• FuzzyART activation_match (#108)
• Info caching (#110)

Merged pull requests:

• Feature/precompilesignatures (#106) (@AP6YC)
• Feature/activation match (#109) (@AP6YC)
• ART statistics (#111) (@AP6YC)
• Release/v0.8.0 (#112) (@AP6YC)

AdaptiveResonance v0.7.0

Diff since v0.6.1

This minor release significantly overhauls the internals of the package in a myriad of ways:

• ElasticArrays are used for the weight growth behavior of modules that benefit from it.
• Activation and match functions are now options that can be swapped out in any module.
• Citations for the package and JOSS paper are added in the readme, documentation, and CITATION.cff file.
• Additional convenience package constants are exported.
• Unit testing time is halved through optimizations and consolidation.
• Unit testing github action pins are updated for the node 12 -> 16 transition.

Closed issues:

• Replace internal types docs with NumericalTypeAliases (#93)
• Remove and consolidate utils (#95)
• Update actions for node 12 -> node 16 (#96)
• Use ElasticArrays.jl (#99)
• Create citation file (#100)

Merged pull requests:

• Feature/doc nta (#94) (@AP6YC)
• Feature/consolidate utils (#97) (@AP6YC)
• Update pins for GitHub actions node update (#98) (@AP6YC)
• Feature/elasticarrays (#101) (@AP6YC)
• Feature/citation (#102) (@AP6YC)
• Bump to v0.7.0 (#103) (@AP6YC)

AdaptiveResonance v0.6.1

Diff since v0.6.0

This patch updates the documentation with additional links and bug fixes.

Closed issues:

• GR bug (#86)
• Logo cleanup (#88)
• Link modules to index (#90)

Merged pull requests:

• Feature/gr bug (#87) (@AP6YC)
• Update logo and header (#89) (@AP6YC)
• Feature/link modules to index (#91) (@AP6YC)
• Release/v0.6.1 (#92) (@AP6YC)

AdaptiveResonance v0.6.0

Diff since v0.5.1

This minor release provides backwards compatibility while significantly expanding documentation, increasing testing coverage, providing speed optimizations with rigorous typing, and more!

Closed issues:

• Codebase idiosyncrasies (#71)
• Update docstrings (#74)
• Add NumericalTypeAliases.jl (#75)
• Concrete members in composite types (#76)
• Fix LFS images in docs (#78)
• Up coverage (#82)

Merged pull requests:

• CompatHelper: bump compat for DocStringExtensions to 0.9, (keep existing compat) (#70) (@github-actions[bot])
• CompatHelper: bump compat for DocStringExtensions to 0.9, (keep existing compat) (#72) (@github-actions[bot])
• Feature/idiosyncrasies (#73) (@AP6YC)
• Add NumericalTypeAliases.jl (#77) (@AP6YC)
• Add with lfs to docs github action (#79) (@AP6YC)
• Concrete members in composite types (#80) (@AP6YC)
• Feature/update docstrings (#81) (@AP6YC)
• Featuer/up coverage (#83) (@AP6YC)
• CompatHelper: bump compat for NumericalTypeAliases to 0.2, (keep existing compat) (#84) (@github-actions[bot])
• v0.6.0 (#85) (@AP6YC)

AdaptiveResonance v0.5.1

Diff since v0.5.0

This patch implements some cosmetic updates, such as:

1. Fixing warnings about incremental compilation being broken during module import.
2. Updating the data loading steps in the DemoCards examples in the documentation.
3. Expanding upon the visualizations in the examples.

Closed issues:

• Zenodo DOI reference (#58)
• CI Optimizations (#60)

Merged pull requests:

• Add zenodo permanent doi to README (#59) (@AP6YC)
• Feature/ci optimization (#61) (@AP6YC)
• Feature/x86 support (#62) (@AP6YC)
• feature/macOS-support (#63) (@AP6YC)
• Exclude 32-bit macOS from CI due to Julia support (#64) (@AP6YC)
• Examples plotting and development (#65) (@AP6YC)
• Remove pyplot backend for docs figures (#66) (@AP6YC)
• Feature/docstring warnings (#67) (@AP6YC)
• Feature/up mldatasets (#68) (@AP6YC)
• Feature/mldatasets fix (#69) (@AP6YC)

AdaptiveResonance v0.5.0

Diff since v0.4.3

This release significantly overhauls the documentation of the package, providing docstrings and an organized package index page in the documentation that describes the hyperparameters and options of all exported names. This version is bumped to a minor release due to new package dependencies.

Closed issues:

• [documentation] DataConfig parameters / Index (#28)

Merged pull requests:

• typos and suggestions (#55) (@aaronpeikert)
• Feature/refine index (#56) (@AP6YC)
• Release/0.5.0 (#57) (@AP6YC)

AdaptiveResonance v0.4.3

Diff since v0.4.2

This release hotfixes citation issues in the attached JOSS file.

AdaptiveResonance v0.4.2

Diff since v0.4.1

This release principally updates the associated JOSS paper with the project for submission, versioned for future reference.

Closed issues:

• [paper] How does AdaptiveResonance.jl compare with other packages? (#23)
• [paper] State of the field (#25)

Merged pull requests:

• Feature/paper review (#51) (@AP6YC)
• Feature/state of field (#52) (@AP6YC)
• ✏️ Fix a couple typos / paper review (#53) (@hayesall)
• Release/0.4.2 (#54) (@AP6YC)

AdaptiveResonance v0.4.1

Diff since v0.4.0

This patch adds a variety of examples in the documentation using DemoCards.jl.

Closed issues:

• [paper] Statement of need - Target audience (#24)
• [documentation] What is the 'Hello World' for this code? (#26)
• [documentation] Minimal working examples (#27)
• [bug?] DDVFA mismatch_flag should probably be false in supervised classification (#42)

Merged pull requests:

• Feature/docs democards (#49) (@AP6YC)
• Release/0.4.1 (#50) (@AP6YC)

AdaptiveResonance v0.4.0

Diff since v0.3.7

This minor release refactors most internal methods to separate incremental and batch methods along with standardizing numerical types across all modules.

Closed issues:

• [documentation,minor] HTML comments in the README can cause rendering problems (#30)
• Warning of method definition (#41)

Merged pull requests:

• Feature/docs additions (#39) (@AP6YC)
• Feature/dataconfig additions (#40) (@AP6YC)
• Feature/fuzzyart (#43) (@AP6YC)
• Feature/ddvfa refactor (#44) (@AP6YC)
• Feature/artmap refactor (#45) (@AP6YC)
• Add target audience, theory, and dynamics to paper (#46) (@AP6YC)
• Add get_bmu sections to docs and README (#47) (@AP6YC)
• Release/0.4 (#48) (@AP6YC)

AdaptiveResonance v0.3.7

Diff since v0.3.6

This patch greatly expands the contribution instructions in the docs and README.

Closed issues:

• dvfa_supervised example header (#17)
• How should users contribute/ask questions? (#18)
• [documentation] Developer Notes (#22)
• [documentation] Community Guidelines (#29)

Merged pull requests:

• Adopt Berlin Code of Conduct (#19) (@AP6YC)
• Fix typo in dvfa_supervised.jl example script (#34) (@AP6YC)
• Developer notes (#35) (@AP6YC)
• Move and expand contributing in README, delete multiline comments (#36) (@AP6YC)
• Fix the default contribution link in README (#37) (@AP6YC)
• Release/0.3.7 (#38) (@AP6YC)

AdaptiveResonance v0.3.6

Diff since v0.3.5

All ART modules can now be instantiated with keyword arguments for their hyperparameters and options in addition to the original default constructor and directly passing the parameters struct.

AdaptiveResonance v0.3.5

Diff since v0.3.4

• Hotfix a DVFA error that normalized against the first weight rather than the current one.

Closed issues:

• Activation function (#16)

AdaptiveResonance v0.3.4

Diff since v0.3.3

Random.jl dependencies have been removed from tests and examples for determinism across all platforms where the same random seed may produce different results.

AdaptiveResonance v0.3.3

Diff since v0.3.2

Hotfix a type edge case where input data loaded as abstract floats for DDVFA would break low-level linear algebra functions.

AdaptiveResonance v0.3.2

Diff since v0.3.1

Remove ClusterValidityIndices.jl requirement.