binary conventions for sparse arrays

src/supervised/supervised.jl

@@ -101,6 +101,34 @@ preallocated buffer, which has to be the same size as the parameters.
     end
 end
 
+##
+# Function for sparse arrays
+@generated function value!{T,N,Q,Ti,M}(
+        buffer::AbstractArray,
+        loss::MarginLoss,
+        target::AbstractSparseArray{Q,Ti,M},
+        output::AbstractArray{T,N}
+    )
+    M > N && throw(ArgumentError("target has more dimensions than output; broadcasting not supported in this direction."))
+    quote
+      @_dimcheck size(buffer) == size(output)
+      @nexprs $M (n)->@_dimcheck(size(target,n) == size(output,n))
+      zeroQ = zero(Q)
+      negQ = Q(-1)
+      @simd for I in CartesianRange(size(output))
+          @nexprs $N n->(i_n = I[n])
+          tgt = @nref($M,target,i)
+          if tgt == zeroQ
+              # convention is that zeros in a sparse array are interpreted as negative one
+              @inbounds @nref($N,buffer,i) = value(loss, negQ, @nref($N,output,i))
+          else
+              @inbounds @nref($N,buffer,i) = value(loss, tgt, @nref($N,output,i))
+          end
+      end
+      buffer
+    end
+end
+
 """
     deriv!(buffer::AbstractArray, loss::SupervisedLoss, target::AbstractArray, output::AbstractArray)
 

test/tst_loss.jl

@@ -393,3 +393,42 @@ end
     end
 end
 
+@testset "Test sparse array conventions for margin-based losses" begin
+    @testset "sparse vector target, vector output" begin
+        N = 50
+
+        # sparse vector of {0,1}
+        sparse_target = sprand(N,0.5)
+        nz = sparse_target .> 0.0
+        sparse_target[nz] = 1.0
+        @test typeof(sparse_target) <: AbstractSparseArray
+
+        # dense vector of {-1,1}
+        target = [ t == 0.0 ? -1.0 : 1.0 for t in sparse_target ]
+
+        output = randn(N)
+
+        for loss in margin_losses
+            @test isapprox(value(loss,sparse_target,output), value(loss,target,output)) 
+        end
+    end
+
+    @testset "sparse vector target, matrix output" begin
+        N = 50
+
+        # sparse vector of {0,1}
+        sparse_target = sprand(N,0.5)
+        nz = sparse_target .> 0.0
+        sparse_target[nz] = 1.0
+        @test typeof(sparse_target) <: AbstractSparseArray
+
+        # dense vector of {-1,1}
+        target = [ t == 0.0 ? -1.0 : 1.0 for t in sparse_target ]
+
+        output = randn(N,N)
+
+        for loss in margin_losses
+            @test isapprox(value(loss,sparse_target,output), value(loss,target,output)) 
+        end
+    end
+end