add learning after episode

tic_tac_toe.py

@@ -25,8 +25,8 @@ def play(player_provider: PlayerProvider, k_max=10000, with_debug=False):
 	player_provider.set_debug(with_debug)
 	for k in range(0, k_max):
 		report(k)
-		players = player_provider.choose()
 		state = create_empty_state()
+		players = player_provider.choose(state)
 		move = 1
 		run = True
 		other_player = players[-1]
@@ -43,14 +43,14 @@ def play(player_provider: PlayerProvider, k_max=10000, with_debug=False):
 				if not has_moved:
 					if with_debug:
 						print(f"{player.mark}: No more moves")
-						player.end_game(0.0)
-						other_player.end_game(0.0)
+						player.end_game(state, 0.0)
+						other_player.end_game(state, 0.0)
 					run = False
 				if player.has_won(state):
 					if with_debug:
 						print(f"{player.mark}: Has won the game")
-						player.end_game(1.0)
-						other_player.end_game(0.0)
+						player.end_game(state, 1.0)
+						other_player.end_game(state, 0.0)
 					run = False
 
 				other_player = player
@@ -60,21 +60,30 @@ def play(player_provider: PlayerProvider, k_max=10000, with_debug=False):
 				move += 1
 
 
-with open("values_x.json", "r") as fp:
-	x_values = json.load(fp)
-
-with open("values_o.json", "r") as fp:
-	o_values = json.load(fp)
-
 do_training = 1
-do_human_player = 1
+do_human_player = 0
+do_load_values = 1
+
+if do_load_values:
+	try:
+		with open("values_x.json", "r") as fp:
+			x_values = json.load(fp)
+
+		with open("values_o.json", "r") as fp:
+			o_values = json.load(fp)
+	except Exception:
+		x_values = {}
+		o_values = {}
+else:
+	x_values = {}
+	o_values = {}
 
 p_hx = HumanPlayer(mark='X', name="Jens")
 p_mx = MachinePlayer(mark='X', params=MachinePlayer.Params(p_explore=0.2, alpha=0.1), values=x_values)
 p_mo = MachinePlayer(mark='O', params=MachinePlayer.Params(p_explore=0.2, alpha=0.1), values=o_values)
 
 if do_training:
-	play(PlayerProvider(p_mx, p_mo), 10000, False)
+	play(PlayerProvider(p_mx, p_mo), 5000, True)
 
 	# Values after training
 	# Convert and write JSON object to file