Optim.jl is not THAT good

Julia/HDS.jl

@@ -60,9 +60,7 @@ function rbf_cost_function(params::Vector{Float64}, centres::Vector{Float32}, ta
     #return plot(predictions, label="Predictions", xlabel="Pixel", ylabel="Intensity", legend=:topleft)
 
     # the cost is the sum of the squared differences between the predictions and the targets
-    rmse = sqrt(sum((predictions .- targets) .^ 2) / length(targets))
-
-    return rmse
+    return sqrt(sum((predictions .- targets) .^ 2))
 end
 
 function rbf_forward(params::Vector{Float64}, centres::Vector{Float32}, targets::Vector{Float32})
@@ -106,7 +104,7 @@ println("#params: ", length(params))
 @time rbf_cost_function(params, centres, row)
 
 # optimize the parameters with Optim
-#result = optimize(x -> rbf_cost_function(x, centres, row), params, Optim.Options(iterations=100, show_trace=true))
+#result = optimize(x -> rbf_cost_function(x, centres, row), params, Optim.Options(iterations=1000000, show_trace=true))
 #result = optimize(x -> rbf_cost_function(x, centres, row), params, LBFGS(), Optim.Options(iterations=1000, show_trace=true))
 result = optimize(x -> rbf_cost_function(x, centres, row), params, SimulatedAnnealing(), Optim.Options(iterations=1000000, show_trace=true))
 println(result)