Merge pull request #578 from alan-turing-institute/bug-576

Fix transposed `y` and `yhat` in measures from LossFunctions
JuliaAI · Jun 10, 2021 · ab2a1ab · ab2a1ab
2 parents 2ef3a07 + 3297993
commit ab2a1ab
Show file tree

Hide file tree

Showing 2 changed files with 6 additions and 6 deletions.
diff --git a/src/measures/loss_functions_interface.jl b/src/measures/loss_functions_interface.jl
@@ -115,7 +115,7 @@ MMI.target_scitype(::Type{<:DistanceLoss}) = Union{Vec{Continuous},Vec{Count}}
 
 function value(measure::DistanceLoss, yhat, X, y, ::Nothing,
                 ::Val{false}, ::Val{true})
-    return LossFunctions.value(getfield(measure, :loss), yhat, y)
+    return LossFunctions.value(getfield(measure, :loss), y, yhat)
 end
 
 function value(measure::DistanceLoss, yhat, X, y, w,
@@ -137,7 +137,7 @@ function value(measure::MarginLoss, yhat, X, y, ::Nothing,
     check_pools(yhat, y)
     probs_of_observed = broadcast(pdf, yhat, y)
     return (LossFunctions.value).(getfield(measure, :loss),
-                                  _scale.(probs_of_observed), 1)
+                                  1, _scale.(probs_of_observed))
 end
 
 function value(measure::MarginLoss, yhat, X, y, w,

diff --git a/test/measures/loss_functions_interface.jl b/test/measures/loss_functions_interface.jl
@@ -36,10 +36,10 @@ const DISTANCE_LOSSES = MLJBase.DISTANCE_LOSSES
     for M_ex in MARGIN_LOSSES
         m = eval(:(MLJBase.$M_ex()))
         @test MLJBase.value(m, yhat, X, y, nothing) ≈
-            LossFunctions.value(getfield(m, :loss), yhatm, ym) ≈
+            LossFunctions.value(getfield(m, :loss), ym, yhatm) ≈
             m(yhat, y)
         @test mean(MLJBase.value(m, yhat, X, y, w)) ≈
-            LossFunctions.value(getfield(m, :loss), yhatm, ym,
+            LossFunctions.value(getfield(m, :loss), ym, yhatm,
                                 LossFunctions.AggMode.WeightedMean(w)) ≈
                                     mean(m(yhat, y, w))
     end
@@ -58,10 +58,10 @@ end
         m_ex = MLJBase.snakecase(M_ex)
         @test m == eval(:(MLJBase.$m_ex))
         @test MLJBase.value(m, yhat, X, y, nothing) ≈
-            LossFunctions.value(getfield(m, :loss), yhat, y) ≈
+            LossFunctions.value(getfield(m, :loss), y, yhat) ≈
             m(yhat, y)
         @test mean(MLJBase.value(m, yhat, X, y, w)) ≈
-            LossFunctions.value(getfield(m, :loss), yhat, y,
+            LossFunctions.value(getfield(m, :loss), y, yhat,
                                 LossFunctions.AggMode.WeightedMean(w)) ≈
                                     mean(m(yhat ,y, w))
     end