use regular +

src/activation.jl

@@ -1,6 +1,6 @@
 export σ, relu, softmax, flatten
 
-σ(x) = 1 ./ (1 .+ exp.(-x))
+σ(x) = 1 ./ (1 + exp.(-x))
 
 back!(::typeof(σ), Δ, x) = Δ .* σ(x)./(1.-σ(x))
 

src/backend/tensorflow/graph.jl

@@ -17,7 +17,7 @@ graph(::typeof(relu), x) = nn.relu(x)
 graph(::typeof(σ), x) = nn.sigmoid(x)
 graph(::typeof(hcat), xs...) = concat(1, xs)
 graph(::typeof(seq), xs, n) = TensorFlow.unpack(xs, num = n, axis = 1)
-graph(::typeof(.+), args...) = +(args...)
+graph(::typeof(+), args...) = +(args...)
 
 for op in (tanh, *, .*, +, -)
   @eval graph(::typeof($op), args...) = $op(node(args)...)

src/layers/affine.jl

@@ -5,7 +5,7 @@ export Affine
 @net type Affine
   W
   b
-  x -> x*W .+ b
+  x -> x*W + b
 end
 
 Affine(in::Integer, out::Integer; init = initn) =

src/layers/recurrent.jl

@@ -36,12 +36,12 @@ GatedRecurrent(in, out; init = initn) =
   y; state
   function (x)
     # Gates
-    forget = σ( x * Wxf .+ y{-1} * Wyf .+ bf )
+    forget = σ( x * Wxf + y{-1} * Wyf + bf )
-    input  = σ( x * Wxi .+ y{-1} * Wyi .+ bi )
+    input  = σ( x * Wxi + y{-1} * Wyi + bi )
-    output = σ( x * Wxo .+ y{-1} * Wyo .+ bo )
+    output = σ( x * Wxo + y{-1} * Wyo + bo )
     # State update and output
-    state′ = tanh( x * Wxc .+ y{-1} * Wyc .+ bc )
+    state′ = tanh( x * Wxc + y{-1} * Wyc + bc )
-    state  = forget .* state{-1} .+ input .* state′
+    state  = forget .* state{-1} + input .* state′
     y = output .* tanh(state)
   end
 end

src/model.jl

@@ -74,7 +74,7 @@ Accumulates the update `Δ` on `p`. The value of `p` won't change until
 `update!`.
 """
 function accumulate!(p::Param, Δ)
-  p.Δx .+= Δ
+  p.Δx += Δ
   return p
 end