Ct and changes to differentiation

CHANGELOG.md

@@ -6,6 +6,35 @@ 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 [v1.1.1].
 
+## Version [1.4.6] - 2026-02-06
+
+### Removed
+
+- Temporarily removed tests for NeuroTreeModels.jl pending updates to compat.
+- Temporarily removed tests for benchmarking, which seems non-performant (will open a separate issue).
+
+### Added
+
+- Added a new evaluation measure `feature_sensitivity(ce::AbstractCounterfactualExplanation)` that can be used to compute the proposed absolute changes for features.
+- Added an optional `callback::Union{Nothing,Function}` argument to `generate_counterfactual` to allow users to either display or store anything relevant from the full `CounterfactualExplanation` object. This is useful if the user has specified `return_flattened=true` but still needs something specific from `CounterfactualExplanation` that is not returned with `FlattenedCE`.
+
+### Changed 
+
+- Moved from Flux.jl for autodiff to DifferentiationInterface.jl. Not all backends working for basic examples, but considerable speedup for ForwardDiff compared to Zygote. The following backends work for generic generator:
+  - Zygote.jl
+  - ForwardDiff.jl
+  - PolyesterForwardDiff.jl
+  - GTPSA.jl
+  - FiniteDiff.jl
+  - Symbolics.jl
+- Avoid aggregating divergence metrics across runs.
+- Changed the way mutability constraints can be supplied: users can now supply a tuple of pairs of feature indices and there corresponding constraints.
+- Removed a bug in `reconstruct_cat_encoding` that turned the `counterfactual_state` object from a matrix into a vector.
+- Added option to provide additional keyword arguments to `validity_strict(ce::CounterfactualExplanation; kwrgs...)`.
+- Small bug fix for `validity` function.
+- Improved `find_potential_neighbours` method.
+- Added possibility to run `compute_divergence` for a fixed number of samples. 
+
 ## Version [1.4.5] - 2025-01-13
 
 ### Changed

Project.toml

@@ -1,16 +1,18 @@
 name = "CounterfactualExplanations"
 uuid = "2f13d31b-18db-44c1-bc43-ebaf2cff0be0"
+version = "1.5.0"
 authors = ["Patrick Altmeyer <p.altmeyer@tudelft.nl> and contributors"]
-version = "1.4.5"
 
 [deps]
 CategoricalArrays = "324d7699-5711-5eae-9e2f-1d82baa6b597"
 CausalInference = "8e462317-f959-576b-b3c1-403f26cec956"
 ChainRulesCore = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4"
 DataFrames = "a93c6f00-e57d-5684-b7b6-d8193f3e46c0"
+DifferentiationInterface = "a0c0ee7d-e4b9-4e03-894e-1c5f64a51d63"
 Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
 EnergySamplers = "f446124b-5d5e-4171-a6dd-a1d99768d3ce"
 Flux = "587475ba-b771-5e3f-ad9e-33799f191a9c"
+ForwardDiff = "f6369f11-7733-5829-9624-2563aa707210"
 Graphs = "86223c79-3864-5bf0-83f7-82e725a168b6"
 KernelFunctions = "ec8451be-7e33-11e9-00cf-bbf324bd1392"
 LazyArtifacts = "4af54fe1-eca0-43a8-85a7-787d91b784e3"
@@ -20,6 +22,7 @@ MLJBase = "a7f614a8-145f-11e9-1d2a-a57a1082229d"
 MLJDecisionTreeInterface = "c6f25543-311c-4c74-83dc-3ea6d1015661"
 MLUtils = "f1d291b0-491e-4a28-83b9-f70985020b54"
 MultivariateStats = "6f286f6a-111f-5878-ab1e-185364afe411"
+Optimisers = "3bd65402-5787-11e9-1adc-39752487f4e2"
 PackageExtensionCompat = "65ce6f38-6b18-4e1d-a461-8949797d7930"
 ProgressMeter = "92933f4c-e287-5a05-a399-4b506db050ca"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
@@ -29,41 +32,42 @@ StatsBase = "2913bbd2-ae8a-5f71-8c99-4fb6c76f3a91"
 Tables = "bd369af6-aec1-5ad0-b16a-f7cc5008161c"
 TaijaBase = "10284c91-9f28-4c9a-abbf-ee43576dfff6"
 UUIDs = "cf7118a7-6976-5b1a-9a39-7adc72f591a4"
+Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"
 
 [weakdeps]
 DecisionTree = "7806a523-6efd-50cb-b5f6-3fa6f1930dbb"
 JointEnergyModels = "48c56d24-211d-4463-bbc0-7a701b291131"
 LaplaceRedux = "c52c1a26-f7c5-402b-80be-ba1e638ad478"
-NeuroTreeModels = "1db4e0a5-a364-4b0c-897c-2bd5a4a3a1f2"
 
 [extensions]
 DecisionTreeExt = "DecisionTree"
 JEMExt = "JointEnergyModels"
 LaplaceReduxExt = "LaplaceRedux"
-NeuroTreeExt = "NeuroTreeModels"
 
 [compat]
 Aqua = "0.8"
-CategoricalArrays = "0.10"
-CausalInference = "0.17, 0.18"
+CategoricalArrays = "0.10, 1"
+CausalInference = "0.19.2"
 ChainRulesCore = "1.15"
 DataFrames = "1"
 DecisionTree = "0.12.3, 0.12.4"
+DifferentiationInterface = "0.7.16"
 Distributions = "0.25.97"
-EnergySamplers = "1.0"
+EnergySamplers = "1"
 Flux = "0.12, 0.13, 0.14, 0.15, 0.16"
+ForwardDiff = "1.3"
 Graphs = "1.11.1"
-JointEnergyModels = "0.1.7"
-KernelFunctions = "0.10.64"
-LaplaceRedux = "0.1.4, 0.2, 1.0"
+JointEnergyModels = "0.1, 0.2"
+KernelFunctions = "0.10"
+LaplaceRedux = "1"
 LazyArtifacts = "1"
 LinearAlgebra = "1"
 Logging = "1"
 MLJBase = "0.21, 1"
 MLJDecisionTreeInterface = "0.4.0"
 MLUtils = "0.2, 0.3, 0.4"
 MultivariateStats = "0.9, 0.10"
-NeuroTreeModels = "1.1.3"
+Optimisers = "0.3, 0.4"
 PackageExtensionCompat = "1"
 ProgressMeter = "1"
 Random = "1"
@@ -74,15 +78,14 @@ Tables = "1"
 TaijaBase = "1"
 Test = "1"
 UUIDs = "1"
+Zygote = "0.6, 0.7"
 julia = "1.10"
 
 [extras]
 Aqua = "4c88cf16-eb10-579e-8560-4a9242c79595"
 DecisionTree = "7806a523-6efd-50cb-b5f6-3fa6f1930dbb"
 JointEnergyModels = "48c56d24-211d-4463-bbc0-7a701b291131"
-LaplaceRedux = "c52c1a26-f7c5-402b-80be-ba1e638ad478"
-NeuroTreeModels = "1db4e0a5-a364-4b0c-897c-2bd5a4a3a1f2"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 
 [targets]
-test = ["Aqua", "DecisionTree", "JointEnergyModels", "LaplaceRedux", "NeuroTreeModels", "Test"]
+test = ["Aqua", "DecisionTree", "JointEnergyModels", "Test"]