39 refactor mljflux interface

CHANGELOG.md

@@ -6,6 +6,43 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 
 *Note*: We try to adhere to these practices as of version [v0.2.1].
 
+## Version [0.3.0] - 2024-06-8
+
+### Changed
+
+- adapted the LaplaceClassification and the LaplaceRegression struct to use the new @mlj_model macro from MLJBase.
+- Changed the fit! method arguments. Now it also accept a Flux chain model instead of retrieving it from the structs. this is due to the fact that MLJ wants only hyperparameters in the struct https://juliaai.github.io/MLJModelInterface.jl/dev/quick_start_guide/
+- Changed the predict functions for both LaplaceClassification and  LaplaceRegression.
+
+### Removed
+- Removed the shape , build and clean! functions.
+
+
+## Version [0.2.3] - 2024-05-31
+
+### Changed
+
+- Removed the link_approx parameter in LaplaceRegression since it is not required.
+- Changed MMI.clean! to check the value of link_approx only in the case likelihood is set to :classification
+- Now the likelihood type in LaplaceClassification and LaplaceRegression is automatically set by the inner constructor. The user is not required to provide it as a parameter anymore.
+
+
+
+
+## Version [0.2.2] - 2024-05-30
+
+### Changed
+
+- Unified duplicated function MMI.clean!: previously MMI.clean! consisted of two separate functions for handling :classification and :regression types respectively. Now, a single MMI.clean! function handles both cases efficiently.[#39]
+- Split LaplaceApproximation struct in two different structs:LaplaceClassification and LaplaceRegression  [#39] 
+- Unified the MLJFlux.shape and the MLJFlux.build functions to handle both :classification and :regression tasks. In particular, shape now handles multi-output regression cases too [[#39](https://github.com/JuliaTrustworthyAI/LaplaceRedux.jl/issues/39)]
+- Changed model metadata for LaplaceClassification and LaplaceRegression
+
+### Added
+ Added Distributions to LaplaceRedux dependency ( needed for MMI.predict(model::LaplaceRegression, fitresult, Xnew) )
+
+
+
 ## Version [0.2.1] - 2024-05-29
 
 ### Changed

Project.toml

@@ -7,8 +7,10 @@ version = "0.2.1"
 ChainRulesCore = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4"
 Compat = "34da2185-b29b-5c13-b0c7-acf172513d20"
 ComputationalResources = "ed09eef8-17a6-5b46-8889-db040fac31e3"
+Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
 Flux = "587475ba-b771-5e3f-ad9e-33799f191a9c"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
+MLJBase = "a7f614a8-145f-11e9-1d2a-a57a1082229d"
 MLJFlux = "094fc8d1-fd35-5302-93ea-dabda2abf845"
 MLJModelInterface = "e80e1ace-859a-464e-9ed9-23947d8ae3ea"
 MLUtils = "f1d291b0-491e-4a28-83b9-f70985020b54"

src/LaplaceRedux.jl

@@ -20,6 +20,6 @@ export optimize_prior!,
     glm_predictive_distribution, posterior_covariance, posterior_precision
 
 include("mlj_flux.jl")
-export LaplaceApproximation
-
+export LaplaceClassification
+export LaplaceRegression
 end