v0.8.1

LIBSVM v0.8.1

Diff since v0.8.0

• update to use LIBLINEAR.jl v0.7

Merged pull requests:

• Add MLJ usage to the README.md (#92) (@ablaom)
• Bump [compat] for LIBLINEAR (#99) (@ablaom)
• Test: cover LIBLINEAR usage (#100) (@barucden)
• Test: serialize models to temporary directory (#101) (@barucden)
• CI: Update CI pipeline (#102) (@barucden)

Closed issues:

• Code in Readme.md does not work. (#90)
• Read datasets from libsvm (#93)
• crash with Julia 1.8.4 under Windows 11 (#95)
• Trained model size depends on the training data size (#96)
• coef0 does not contain the intercept (#97)
• Bug? (#98)
• ERROR: Package LIBSVM errored during testing (exit code: 3221226356) (#103)
• Why is LIBSVM superior? (May I ask a slightly off-topic question here?) (#104)

v0.8.0

LIBSVM v0.8.0

Diff since v0.7.0

Closed issues:

• ROC support (#61)
• Using Kernels.Precomputed, svmpredict doesn't verify the dimensions of the input (#85)
• Allow callable as kernel argument (#86)

Merged pull requests:

• Check dimension of gram matrix (#87) (@till-m)
• Callable kernel (#88) (@till-m)

v0.7.0

LIBSVM v0.7.0

Diff since v0.6.0

Closed issues:

• Small improvement (#46)
• Error building LIBSVM (#62)
• How does one specify a pre-computed kernel? (#71)

Merged pull requests:

• remove global variable verbosity (#68) (@iblis17)
• add libsvm_version c global var (#69) (@iblis17)
• Avoid abstract types of fields in SupportVectors (#70) (@barucden)
• Call svm_check_parameter before training (#72) (@barucden)
• Add support for precomputed kernel. (#73) (@barucden)
• LibSVMtypes: using C style type alias (#74) (@iblis17)
• refine check_parameter, use proper type mapping (#75) (@iblis17)
• Wrap ccalls with proper Julia functions (#76) (@barucden)
• refine svmpredict (#77) (@iblis17)
• Eliminate duplicated allocations of at-cfunction for libsvm_set_verbose (#78) (@iblis17)
• Add GC.preserve to extend the object lifecycle (#79) (@iblis17)
• cleanup duplicated libsvm_check_parameter call (#82) (@iblis17)
• README: documentation for precomputed kernel (#83) (@barucden)

v0.6.0

LIBSVM v0.6.0

Diff since v0.5.0

Closed issues:

• Support LIBLinear (#1)
• Missing dependency on Windows (#45)
• Update to allow Compat = 3.0 in REQUIRE or, better, upgrade to Julia 1.0 (#48)
• LIBSVM.jl webpage : example not working (#50)
• Cannot install LIBSVM (#52)
• problem with example (#55)
• Error in Readme example (#57)
• Can't build LIBSVM on Appveyor (#59)

Merged pull requests:

• pre-built libsvm binaries using BinaryBuilder (#53) (@aviks)
• Changed 'polynomial' to 'Polynomial' in docstring (#54) (@MartinuzziFrancesco)
• Update the usage in README. (#58) (@barucden)
• migrate CI to GitHub Actions (#64) (@iblis17)
• Relax label vector type to AbstractVector (#65) (@iblis17)
• CompatHelper: bump compat for "LIBLINEAR" to "0.6" (#66) (@github-actions[bot])

v0.5.0

pre-built libsvm binaries using BinaryBuilder (#53)

* pre-built libsvm binaries using BinaryBuilder

* Minimum julia version is now 1.3

LIBSVM v0.4.0

Add Project.toml file and delete REQUIRE file (#47)

LIBSVM v0.3.1

• Fix deprecation with probabilities

LIBSVM v0.3.0

• Update to julia 0.7, drop 0.6 support

LIBSVM v0.2.0

• Drop 0.5 support
• Add multi-threading
• Fixes to docs

LIBSVM v0.1.1

• Fix ScikitLearn.jl compatibility #31