feat(demo): do not display the regularization part of the loss

SDeMo/src/classifiers/logistic.jl

@@ -4,7 +4,7 @@ function __sigmoid!(store::Vector{<:AbstractFloat}, z::Vector{<:AbstractFloat})
         store[i] = __sigmoid(z[i])
     end
     return store
-end    
+end
 
 @testitem "We get the correct response for a sigmoid" begin
     @test SDeMo.__sigmoid(0.0) == 0.5
@@ -85,7 +85,12 @@ end
 
 Base.zero(::Type{Logistic}) = 0.5
 
-function SDeMo.train!(lreg::Logistic, y::Vector{Bool}, X::Matrix{T}; kwargs...) where {T <: Number}
+function SDeMo.train!(
+    lreg::Logistic,
+    y::Vector{Bool},
+    X::Matrix{T};
+    kwargs...,
+) where {T <: Number}
     # Get the validation data if relevant
     Xt = get(kwargs, :Xt, nothing)
     yt = get(kwargs, :yt, nothing)
@@ -108,12 +113,10 @@ function SDeMo.train!(lreg::Logistic, y::Vector{Bool}, X::Matrix{T}; kwargs...)
             if validation_data
                 zv = Xvt * lreg.θ
                 __sigmoid!(zv, zv)
-                validation_loss = -mean(yt .* log.(zv) .+ (1 .- yt) .* log.(1 .- zv)) + (lreg.λ / (2 * length(lreg.θ))) * sum(lreg.θ[2:end] .^ 2)
+                validation_loss = -mean(yt .* log.(zv) .+ (1 .- yt) .* log.(1 .- zv))
             end
             z = clamp.(z, eps(), 1 - eps())
-            loss =
-                -mean(y .* log.(z) .+ (1 .- y) .* log.(1 .- z)) +
-                (lreg.λ / (2 * length(lreg.θ))) * sum(lreg.θ[2:end] .^ 2)
+            loss = -mean(y .* log.(z) .+ (1 .- y) .* log.(1 .- z))
             # Percent done
             prct = lpad(round(Int64, (epoch / lreg.epochs) * 100), 3, " ")
             infostr = "[$(prct)%] LOSS: training ≈ $(rpad(round(loss; digits=4), 6, " "))"
@@ -205,17 +208,17 @@ end
 
 @testitem "We can run a Logistic model" begin
     X, y = SDeMo.__demodata()
-    sdm = SDM(ZScore(), Logistic(), 0.5, X, y, [1,2,12])
+    sdm = SDM(ZScore(), Logistic(), 0.5, X, y, [1, 2, 12])
     folds = holdout(sdm)
     classifier(sdm).verbose = true
     classifier(sdm).η = 1e-3
     classifier(sdm).verbosity = 10
-    train!(sdm; training=folds[1])
+    train!(sdm; training = folds[1])
 end
 
 @testitem "We can run a verbose Logistic model with no training data" begin
     X, y = SDeMo.__demodata()
-    sdm = SDM(ZScore(), Logistic(), 0.5, X, y, [1,2,12])
+    sdm = SDM(ZScore(), Logistic(), 0.5, X, y, [1, 2, 12])
     classifier(sdm).verbose = true
     classifier(sdm).η = 1e-3
     classifier(sdm).verbosity = 10