using ConformalPrediction using CounterfactualExplanations.Models using Flux using MLJBase using MLUtils using SliceMap using Statistics """ ConformalModel <: Models.AbstractDifferentiableJuliaModel Constructor for models trained in `Flux.jl`. """ struct ConformalModel <: Models.AbstractDifferentiableJuliaModel model::ConformalPrediction.ConformalProbabilisticSet fitresult::Any likelihood::Union{Nothing,Symbol} function ConformalModel(model, fitresult, likelihood) if likelihood ∈ [:classification_binary, :classification_multi] || isnothing(likelihood) new(model, fitresult, likelihood) else throw( ArgumentError( "`likelihood` should either be `nothing` or in `[:classification_binary,:classification_multi]`", ), ) end end end # Outer constructor method: function ConformalModel(model, fitresult=nothing; likelihood::Union{Nothing,Symbol}=nothing) # Check if model is fitted and infer likelihood: if isnothing(fitresult) @info "Conformal Model is not fitted." else outdim = length(fitresult[2]) _likelihood = outdim == 2 ? :classification_binary : :classification_multi @assert likelihood == _likelihood || isnothing(likelihood) "Specification of `likelihood` does not match the output dimension of the model." likelihood = _likelihood end # Default to binary classification, if not specified or inferred: if isnothing(likelihood) likelihood = :classification_binary @info "Likelihood not specified. Defaulting to $likelihood." end # Construct model: M = ConformalModel(model, fitresult, likelihood) return M end # Methods @doc raw""" Models.logits(M::ConformalModel, X::AbstractArray) To keep things consistent with the architecture of `CounterfactualExplanations.jl`, this method computes logits $\beta_i x_i$ (i.e. the linear predictions) for a Conformal Classifier. By default, `MLJ.jl` and `ConformalPrediction.jl` return probabilistic predictions. To get the underlying logits, we invert the softmax function. Let $\hat{p}_i$ denote the estimated softmax output for feature $i$. Then in the multi-class case the following formula can be applied: ```math \beta_i x_i = \log (\hat{p}_i) + \log (\sum_i \exp(\hat{p}_i)) ``` For a short derivation, see here: https://math.stackexchange.com/questions/2786600/invert-the-softmax-function. In the binary case logits are fed through the sigmoid function instead of softmax, so we need to further adjust as follows, ```math \beta x = \beta_1 x_1 - \beta_0 x_0 ``` which follows from the derivation here: https://stats.stackexchange.com/questions/233658/softmax-vs-sigmoid-function-in-logistic-classifier """ function Models.logits(M::ConformalModel, X::AbstractArray) fitresult = M.fitresult function predict_logits(fitresult, x) p̂ = fitresult[1](x) if ndims(p̂) == 2 p̂ = [p̂] end p̂ = reduce(hcat, p̂) ŷ = reduce(hcat, (map(p -> log.(p) .+ log(sum(exp.(p))), eachcol(p̂)))) if M.likelihood == :classification_binary ŷ = reduce(hcat, (map(y -> y[2] - y[1], eachcol(ŷ)))) end ŷ = ndims(ŷ) > 1 ? ŷ : permutedims([ŷ]) return ŷ end if ndims(X) > 2 yhat = map(eachslice(X, dims=ndims(X))) do x predict_logits(fitresult, x) end yhat = MLUtils.stack(yhat) else yhat = predict_logits(fitresult, X) end return yhat end """ Models.probs(M::ConformalModel, X::AbstractArray) Returns the estimated probabilities for a Conformal Classifier. """ function Models.probs(M::ConformalModel, X::AbstractArray) if M.likelihood == :classification_binary output = σ.(Models.logits(M, X)) elseif M.likelihood == :classification_multi output = softmax(Models.logits(M, X)) end return output end """ train(M::ConformalModel, data::CounterfactualData; kwrgs...) Trains a Conformal Classifier `M` on `data`. """ function Models.train(M::ConformalModel, data::CounterfactualData; kwrgs...) X, y = data.X, data.output_encoder.labels X = table(permutedims(X)) conf_model = M.model mach = machine(conf_model, X, y) fit!(mach; kwrgs...) likelihood, _ = CounterfactualExplanations.guess_likelihood(data.output_encoder.y) return ConformalModel(mach.model, mach.fitresult, likelihood) end