Merge f494748 into 50e27ea

src/optimization/convDual.jl

@@ -82,6 +82,7 @@ function get_bounds(nnet::Network, input::Vector{Float64}, ϵ::Float64)
     u = Vector{Vector{Float64}}()
     γ = Vector{Vector{Float64}}()
     μ = Vector{Vector{Vector{Float64}}}()
+    input_ReLU = Vector{Vector{Float64}}()
 
     v1 = layers[1].weights'
     push!(γ, layers[1].bias)
@@ -94,19 +95,23 @@ function get_bounds(nnet::Network, input::Vector{Float64}, ϵ::Float64)
         n_input  = length(layers[i-1].bias)
         n_output = length(layers[i].bias)
 
-        input_ReLU = relaxed_ReLU.(last(l), last(u))
-        D = Diagonal(input_ReLU)   # a matrix whose diagonal values are the relaxed_ReLU values (maybe should be sparse?)
+        last_input_ReLU = relaxed_ReLU.(last(l), last(u))
+        push!(input_ReLU, last_input_ReLU)
+        D = Diagonal(last_input_ReLU)   # a matrix whose diagonal values are the relaxed_ReLU values (maybe should be sparse?)
 
         # Propagate existing terms
         WD = layers[i].weights*D
         v1 = v1 * WD' # TODO CHECK
         map!(g -> WD*g,   γ, γ)
+
+        # Updating ν_j for all previous layers
         for M in μ
             map!(m -> WD*m,   M, M)
         end
+
         # New terms
         push!(γ, layers[i].bias)
-        push!(μ, new_μ(n_input, n_output, input_ReLU, WD))
+        push!(μ, new_μ(n_input, n_output, last_input_ReLU, WD))
 
         # Compute bounds
         ψ = v1' * input + sum(γ)
@@ -127,7 +132,7 @@ function all_neg_pos_sums(slopes, l, μ, n_output)
     # Need to debug
     for (i, ℓ) in enumerate(l)                # ℓ::Vector{Float64}
         for (j, M) in enumerate(μ[i])         # M::Vector{Float64}
-            if 0 < slopes[j] < 1              # if in the triangle region of relaxed ReLU
+            if 0 < slopes[i][j] < 1              # if in the triangle region of relaxed ReLU
                 #posind = M .> 0
                 neg .+= ℓ[j] * min.(M, 0) #-M .* !posind  # multiply by boolean to set the undesired values to 0.0
                 pos .+= ℓ[j] * max.(M, 0) #M .* posind