- simplified debug flags

- added params class
- cleaned up

a_player.py

@@ -11,6 +11,9 @@ class APlayer(object):
         self.state = None
         self.state_last = None
 
+    def set_debug(self, with_debug):
+        pass
+
     def move(self, state: np.array):
         return state, False
 

machine_player.py

@@ -5,15 +5,22 @@ from numpy.random import uniform
 
 
 class MachinePlayer(APlayer):
-    def __init__(self, mark='X', with_debug=False, values=None):
+    class Params:
+        def __init__(self, p_exp, alpha):
+            self.p_exp = p_exp
+            self.alpha = alpha
+
+    def __init__(self, mark, params: Params):
         APlayer.__init__(self, mark)
-        self.p_exp = 0.2
+        self.params = params
-        self.alpha = 0.1
+        self.values = {}
+        self.with_debug = False
+
+    def set_debug(self, with_debug):
         self.with_debug = with_debug
-        if values is None:
+
-            self.init_values()
+    def assign_values(self, values):
-        else:
+        self.values = values
-            self.values = values
 
     def init_values(self):
         self.values = {}
@@ -78,7 +85,7 @@ class MachinePlayer(APlayer):
 
         next_move = best_move
         is_exp = False
-        if uniform() < self.p_exp:
+        if uniform() < self.params.p_exp:
             index = np.random.randint(len(moves))
             next_move = moves[index]
             is_exp = best_move != next_move
@@ -101,7 +108,7 @@ class MachinePlayer(APlayer):
         if not is_exp and self.state_last is not None:
             v0 = self.get_value(self.state_last)
             v1 = self.get_value(self.state)
-            d = v0 + self.alpha*(v1-v0)
+            d = v0 + self.params.alpha*(v1-v0)
             if d > 0:
                 self.set_value(self.state_last, d)
             if self.with_debug:

player_provider.py

@@ -6,6 +6,10 @@ class PlayerProvider:
     def __init__(self, p1: APlayer, p2: APlayer):
         self.players = [p1, p2]
 
+    def set_debug(self, with_debug):
+        for p in self.players:
+            p.set_debug(with_debug)
+
     def choose(self) -> list[APlayer]:
         # Wer fängt an?
         players = self.players

tic_tac_toe.py

@@ -8,8 +8,8 @@ float_formatter = "{:.3f}".format
 np.set_printoptions(formatter={'float_kind': float_formatter})
 
 
-
+def play(player_provider: PlayerProvider, k_max=10000, with_debug=False):
-def play(player_provider: PlayerProvider, k_max=10000, with_print=False):
+    player_provider.set_debug(with_debug)
     for k in range(0, k_max):
         players = player_provider.choose()
         state = create_empty_state()
@@ -18,16 +18,16 @@ def play(player_provider: PlayerProvider, k_max=10000, with_print=False):
         other_player = players[-1]
         while run:
             for player in players:
-                if with_print:
+                if with_debug:
                     print(f"---------------------------------------------------")
                     print(f"- Game {k:06d}, Move {move} -----------------------------")
                     print(f"---------------------------------------------------")
                 last_state = state
                 state, has_moved = player.move(state)
-                if with_print:
+                if with_debug:
                     print(to_state_string(last_state, state))
                 if not has_moved:
-                    if with_print:
+                    if with_debug:
                         print(f"{player.mark}: No more moves")
                     if isinstance(player, MachinePlayer):
                         player.reward(0.0)
@@ -35,7 +35,7 @@ def play(player_provider: PlayerProvider, k_max=10000, with_print=False):
                         other_player.reward(0.0)
                     run = False
                 if player.has_won(state):
-                    if with_print:
+                    if with_debug:
                         print(f"{player.mark}: Has won the game")
                     if isinstance(player, MachinePlayer):
                         player.reward(1.0)
@@ -56,16 +56,19 @@ with open("values_x.json", "r") as fp:
 with open("values_o.json", "r") as fp:
     o_values = json.load(fp)
 
-px = MachinePlayer(mark='X', with_debug=False, values=x_values)
+px = MachinePlayer(mark='X', params=MachinePlayer.Params(p_exp=0.2, alpha=0.1))
-po = MachinePlayer(mark='O', with_debug=False, values=o_values)
+po = MachinePlayer(mark='O', params=MachinePlayer.Params(p_exp=0.2, alpha=0.1))
+
+px.assign_values(x_values)
+po.assign_values(o_values)
+
+player_provider = PlayerProvider(px, po)
 
 do_training = 1
 if do_training:
-    players = PlayerProvider(px, po)
+    play(player_provider, 10000, False)
-    play(players, 10000, False)
 
-players = PlayerProvider(MachinePlayer(mark='X', with_debug=True, values=px.values), MachinePlayer(mark='O', with_debug=True, values=po.values))
+play(player_provider, 10, True)
-play(players, 1000, True)
 
 # Convert and write JSON object to file
 with open("values_x.json", "w") as fp: