Fixed loss function.

julia_code/BellmanResidualMinimization.jl

@@ -218,7 +218,7 @@ function Ξ(
 	bellman_residuals = value((s,d)) - payoff_fn(s,d,a,econ_params) - econ_params.β*value((s′,d′))
 	maximization_condition = - payoff_fn(s,d,a,econ_params) - econ_params.β*value((s′,d′))
 
-	return bellman_residuals.^2 - maximization_condition
+	return sum([bellman_residuals.^2 maximization_condition])
 end
 
 # ╔═╡ a20959be-65e4-4b69-9521-503bc59f0854
@@ -229,18 +229,25 @@ end
 
 # ╔═╡ dff642d9-ec5a-4fed-a059-6c07760a3a58
 #loss function
-loss(s,d) = Flux.mse(Ξ(s,d,bm,em,payoff1),0)
+loss(s,d) = Ξ(s,d,bm,em,payoff1)
+
+# ╔═╡ 20c777b5-4295-4478-8f53-b18cd409c8ae
+s1 = ones(N_constellations)
+
+# ╔═╡ 1d65707c-6333-4252-ace6-bad47146ba06
+d1 = ones(N_debris)
+
+# ╔═╡ 43b99708-0052-4b78-886c-92ac2b532f29
+Ξ(s1,d1,bm,em,payoff1)
+
+# ╔═╡ 39433c1a-c3ac-45b0-b1bf-ff2d42ca9cbb
+
 
 # ╔═╡ 6bf8d29a-7990-4e91-86e6-d9894ed3db27
 #optimizer
 ADAM = Flux.Optimise.ADAM()
 
 # ╔═╡ 74f5fde3-0593-46fc-a688-f1db7ab28c64
-for i in 1:10
-	#print loss under certain conditions 
-	#loss(s1,d1)
-	
-
 for epoch in 1:200
 	#train the policy funciton
 	Flux.Optimise.train!(loss, params(policy), data, ADAM)
@@ -250,7 +257,6 @@ for i in 1:10
 	#Train the value functions
 	Flux.Optimise.train!(loss, params(value), data, ADAM)
 end
-end
 
 # ╔═╡ 02f3fe78-e7a7-453f-9ddf-acddf08d8676
 begin
@@ -849,6 +855,10 @@ uuid = "3f19e933-33d8-53b3-aaab-bd5110c3b7a0"
 # ╠═41271ab4-1ec7-431f-9efb-0f7c3da2d8b4
 # ╠═a20959be-65e4-4b69-9521-503bc59f0854
 # ╠═dff642d9-ec5a-4fed-a059-6c07760a3a58
+# ╠═20c777b5-4295-4478-8f53-b18cd409c8ae
+# ╠═1d65707c-6333-4252-ace6-bad47146ba06
+# ╠═43b99708-0052-4b78-886c-92ac2b532f29
+# ╠═39433c1a-c3ac-45b0-b1bf-ff2d42ca9cbb
 # ╠═6bf8d29a-7990-4e91-86e6-d9894ed3db27
 # ╠═74f5fde3-0593-46fc-a688-f1db7ab28c64
 # ╠═02f3fe78-e7a7-453f-9ddf-acddf08d8676