Update XAIBase dependency to v4, drop support for Julia <1.10

Project.toml

@@ -1,7 +1,7 @@
 name = "RelevancePropagation"
 uuid = "0be6dd02-ae9e-43eb-b318-c6e81d6890d8"
 authors = ["Adrian Hill <[email protected]>"]
-version = "2.0.3-DEV"
+version = "3.0.0-DEV"
 
 [deps]
 Flux = "587475ba-b771-5e3f-ad9e-33799f191a9c"
@@ -14,12 +14,12 @@ XAIBase = "9b48221d-a747-4c1b-9860-46a1d8ba24a7"
 Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"
 
 [compat]
-Flux = "0.13, 0.14"
+Flux = "0.14"
 MacroTools = "0.5"
 Markdown = "1"
 Random = "1"
 Reexport = "1"
 Statistics = "1"
-XAIBase = "3"
+XAIBase = "4"
 Zygote = "0.6"
-julia = "1.6"
+julia = "1.10"

README.md

@@ -12,7 +12,7 @@ This package is part of the [Julia-XAI ecosystem](https://github.com/Julia-XAI)
 [ExplainableAI.jl](https://github.com/Julia-XAI/ExplainableAI.jl).
 
 ## Installation 
-This package supports Julia ≥1.6. To install it, open the Julia REPL and run 
+This package supports Julia ≥1.10. To install it, open the Julia REPL and run 
 ```julia-repl
 julia> ]add RelevancePropagation
 ```

src/RelevancePropagation.jl

@@ -2,6 +2,7 @@ module RelevancePropagation
 
 using Reexport
 @reexport using XAIBase
+import XAIBase: call_analyzer
 
 using XAIBase: AbstractFeatureSelector, number_of_features
 using Base.Iterators
@@ -12,7 +13,6 @@ using Zygote
 using Markdown
 using Statistics: mean, std
 
-include("compat.jl")
 include("bibliography.jl")
 include("layer_types.jl")
 include("layer_utils.jl")

src/crp.jl

@@ -32,7 +32,9 @@ end
 # Call to CRP analyzer #
 #======================#
 
-function (crp::CRP)(input::AbstractArray{T,N}, ns::AbstractOutputSelector) where {T,N}
+function call_analyzer(
+    input::AbstractArray{T,N}, crp::CRP, ns::AbstractOutputSelector
+) where {T,N}
     rules = crp.lrp.rules
     layers = crp.lrp.model.layers
     modified_layers = crp.lrp.modified_layers
@@ -88,5 +90,5 @@ function (crp::CRP)(input::AbstractArray{T,N}, ns::AbstractOutputSelector) where
             end
         end
     end
-    return Explanation(R_return, last(as), ns(last(as)), :CRP, :attribution, nothing)
+    return Explanation(R_return, input, last(as), ns(last(as)), :CRP, :attribution, nothing)
 end

src/lrp.jl

@@ -55,16 +55,16 @@ LRP(model::Chain, c::Composite; kwargs...) = LRP(model, lrp_rules(model, c); kwa
 # Call to the LRP analyzer #
 #==========================#
 
-function (lrp::LRP)(
-    input::AbstractArray, ns::AbstractOutputSelector; layerwise_relevances=false
+function call_analyzer(
+    input::AbstractArray, lrp::LRP, ns::AbstractOutputSelector; layerwise_relevances=false
 )
     as = get_activations(lrp.model, input)    # compute activations aᵏ for all layers k
     Rs = similar.(as)                         # allocate relevances Rᵏ for all layers k
     mask_output_neuron!(Rs[end], as[end], ns) # compute relevance Rᴺ of output layer N
 
     lrp_backward_pass!(Rs, as, lrp.rules, lrp.model, lrp.modified_layers)
     extras = layerwise_relevances ? (layerwise_relevances=Rs,) : nothing
-    return Explanation(first(Rs), last(as), ns(last(as)), :LRP, :attribution, extras)
+    return Explanation(first(Rs), input, last(as), ns(last(as)), :LRP, :attribution, extras)
 end
 
 get_activations(model, input) = (input, Flux.activations(model, input)...)

test/test_batches.jl

@@ -30,20 +30,12 @@ ANALYZERS = Dict(
 
 for (name, method) in ANALYZERS
     @testset "$name" begin
-        # Using `add_batch_dim=true` should result in same explanation
-        # as input reshaped to have a batch dimension
-        analyzer = method(model)
-        expl1_no_bd = analyzer(input1_no_bd; add_batch_dim=true)
-        analyzer = method(model)
-        expl1_bd = analyzer(input1_bd)
-        @test expl1_bd.val ≈ expl1_no_bd.val
-
         # Analyzing a batch should have the same result
         # as analyzing inputs in batch individually
         analyzer = method(model)
         expl2_bd = analyzer(input2_bd)
         analyzer = method(model)
         expl_batch = analyzer(input_batch)
-        @test expl1_bd.val ≈ expl_batch.val[:, 1]
+        @test expl2_bd.val ≈ expl_batch.val[:, 2]
     end
 end