Refactored

helper.py

@@ -2,14 +2,6 @@ import numpy as np
 from numpy.random import uniform
 
 
-def create_test_state() -> np.array:
-	return np.array([['1', '2', '3'], ['4', '5', '6'], ['7', '8', '9']])
-
-
-def create_empty_state() -> np.array:
-	return np.array([['-', '-', '-'], ['-', '-', '-'], ['-', '-', '-']])
-
-
 def softmax(values: np.array, eps=1.0E-6) -> np.array:
 	result = (values + eps) / (sum(values + eps))
 	return result
@@ -46,36 +38,6 @@ def test_sample(data):
 	print(c)
 
 
-def to_state_string(state, state_nex=None):
-	sp = '      '
-	sp_arrow = '  =>  '
-	sp_ = [sp, sp_arrow, sp]
-
-	def col(str_in, state):
-		str_out = str_in
-		for c in range(0, 3):
-			char = state[r][c]
-			if char == '-':
-				char = ' '
-			str_out += "|" + char
-		str_out += '|'
-		return str_out
-
-	state_str = ''
-	for r in range(0, 3):
-		if state is not None:
-			state_str = col(state_str, state)
-
-		if state_nex is not None:
-			state_str += sp_[r]
-			state_str = col(state_str, state_nex)
-
-		if r != 2:
-			state_str += '\x0A'
-
-	return state_str
-
-
 if __name__ == '__main__':
 	print("Testing sample()")
 	test_sample([0.7, 0.1, 0.1, 0.1])

human_player.py

@@ -1,7 +1,7 @@
 import numpy as np
 
 from a_player import APlayer
-from helper import to_state_string, create_empty_state, create_test_state
+from state import to_state_string, create_empty_state, create_test_state
 from state import get_potential_moves
 
 
@@ -17,7 +17,7 @@ class HumanPlayer(APlayer):
 			print(f"{self.name}, Du bist dran. Das Spielfeld sieht so aus:")
 			print(to_state_string(state, create_test_state()))
 
-			print(f"{self.name}, in welches Feld willst Du Dein \"{self.mark}\" setzen?")
+			print(f"{self.name}, in welches Feld willst Du das \"{self.mark}\" setzen?")
 			choice = input("Feldnummer: ")
 
 			if choice in "Qq":
@@ -35,7 +35,7 @@ class HumanPlayer(APlayer):
 				print("Versuche es nochmal")
 				continue
 
-			print("So soll es geschehen")
+			print(f"Okay, nächstes \"{self.mark}\" auf Feld {move+1}")
 			has_moved = True
 			break
 

machine_player.py

@@ -6,8 +6,8 @@ from state import get_potential_moves
 
 class MachinePlayer(APlayer):
 	class Params:
-		def __init__(self, p_exp, alpha):
+		def __init__(self, p_explore, alpha):
-			self.p_exp = p_exp
+			self.p_exp = p_explore
 			self.alpha = alpha
 
 	def __init__(self, mark, params: Params, values):
@@ -53,7 +53,6 @@ class MachinePlayer(APlayer):
 		self.state_last = None
 
 	def move(self, state: np.array):
-		do_sample = False
 		values = np.array([])
 		# get possible move
 		moves = get_potential_moves(state)
@@ -74,13 +73,11 @@ class MachinePlayer(APlayer):
 
 		next_move = best_move
 		is_exp = False
+		# Randomly perform exploratory move
 		if uniform() < self.params.p_exp:
 			index = np.random.randint(len(moves))
 			next_move = moves[index]
 			is_exp = best_move != next_move
-		elif do_sample:
-			index = sample(values)
-			next_move = moves[index]
 
 		if self.with_debug:
 			print(f"{self.mark}: Values    = {values}")

state.py

@@ -1,6 +1,44 @@
 import numpy as np
 
 
+def create_test_state() -> np.array:
+	return np.array([['1', '2', '3'], ['4', '5', '6'], ['7', '8', '9']])
+
+
+def create_empty_state() -> np.array:
+	return np.array([['-', '-', '-'], ['-', '-', '-'], ['-', '-', '-']])
+
+
+def to_state_string(state, state_nex=None):
+	sp = '      '
+	sp_arrow = '  =>  '
+	sp_ = [sp, sp_arrow, sp]
+
+	def col(str_in, state):
+		str_out = str_in
+		for c in range(0, 3):
+			char = state[r][c]
+			if char == '-':
+				char = ' '
+			str_out += "|" + char
+		str_out += '|'
+		return str_out
+
+	state_str = ''
+	for r in range(0, 3):
+		if state is not None:
+			state_str = col(state_str, state)
+
+		if state_nex is not None:
+			state_str += sp_[r]
+			state_str = col(state_str, state_nex)
+
+		if r != 2:
+			state_str += '\x0A'
+
+	return state_str
+
+
 def get_potential_moves(state: np.array) -> np.array:
 	st = state.reshape(state.size)
 	indices = [idx for idx, s in enumerate(st) if '-' in s]

tic_tac_toe.py

@@ -1,5 +1,5 @@
 import numpy as np
-from helper import create_empty_state, to_state_string
+from state import create_empty_state, to_state_string
 from machine_player import MachinePlayer
 from human_player import HumanPlayer
 from player_provider import PlayerProvider
@@ -66,21 +66,28 @@ 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', params=MachinePlayer.Params(p_exp=0.2, alpha=0.1), values=x_values)
-px = HumanPlayer(mark='X', name="Jens")
-po = MachinePlayer(mark='O', params=MachinePlayer.Params(p_exp=0.2, alpha=0.1), values=o_values)
-
-player_provider = PlayerProvider(px, po)
-
 do_training = 1
+do_human_player = 1
+
+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(player_provider, 10000, True)
+	play(PlayerProvider(p_mx, p_mo), 10000, False)
-
-play(player_provider, 10, True)
 
+	# Values after training
 	# Convert and write JSON object to file
 	with open("values_x.json", "w") as fp:
-	json.dump(x_values, fp, indent=0)
+		json.dump(p_mx.values, fp, indent=0)
 
 	with open("values_o.json", "w") as fp:
-	json.dump(o_values, fp, indent=0)
+		json.dump(p_mo.values, fp, indent=0)
+
+if do_human_player:
+	pp = PlayerProvider(p_hx, p_mo)
+else:
+	pp = PlayerProvider(p_mx, p_mo)
+
+play(pp, 10, True)
+