-
Notifications
You must be signed in to change notification settings - Fork 0
/
agent.py
132 lines (111 loc) · 3.35 KB
/
agent.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
from board import Board
import numpy as np
import random
import os
class Agent():
def __init__(self):
self.episode = []
self.net = []
self.alpha = 0.0025
self.gamma = 1.0
if os.path.isfile("tupleNet/tuple1.npy"):
print("Found tuple network")
print("Loading...")
self.load_tupleNet("tupleNet/tuple")
else:
print("Building tuple Network...")
self.build_tupleNet()
print("Completed")
def build_tupleNet(self):
self.net.append(np.zeros(shape=(20, 20, 20, 20, 20, 20), dtype=np.float32))
self.net.append(np.zeros(shape=(20, 20, 20, 20, 20, 20), dtype=np.float32))
self.net.append(np.zeros(shape=(20, 20, 20, 20), dtype=np.float32))
self.net.append(np.zeros(shape=(20, 20, 20, 20), dtype=np.float32))
def load_tupleNet(self, filename):
for i in range(4):
self.net.append(np.load(filename+str(i+1)+".npy"))
def save_tupleNet(self):
for i in range(4):
np.save("tupleNet/tuple%d" % (i+1), self.net[i])
def updateNet(self, tmp, TD_error):
self.net[0][tmp.getTile(0)][tmp.getTile(4)][tmp.getTile(8)][tmp.getTile(1)][tmp.getTile(5)][tmp.getTile(9)] += TD_error
self.net[1][tmp.getTile(1)][tmp.getTile(5)][tmp.getTile(9)][tmp.getTile(2)][tmp.getTile(6)][tmp.getTile(10)] += TD_error
self.net[2][tmp.getTile(2)][tmp.getTile(6)][tmp.getTile(10)][tmp.getTile(14)] += TD_error
self.net[3][tmp.getTile(3)][tmp.getTile(7)][tmp.getTile(11)][tmp.getTile(15)] += TD_error
def getV(self, b):
v = 0.0
tmp = Board()
for i in range(8):
tmp.copyBoard(b)
tmp.morphBoard(i)
v += self.net[0][tmp.getTile(0)][tmp.getTile(4)][tmp.getTile(8)][tmp.getTile(1)][tmp.getTile(5)][tmp.getTile(9)]
v += self.net[1][tmp.getTile(1)][tmp.getTile(5)][tmp.getTile(9)][tmp.getTile(2)][tmp.getTile(6)][tmp.getTile(10)]
v += self.net[2][tmp.getTile(2)][tmp.getTile(6)][tmp.getTile(10)][tmp.getTile(14)]
v += self.net[3][tmp.getTile(3)][tmp.getTile(7)][tmp.getTile(11)][tmp.getTile(15)]
return v
def Episode_begin(self):
self.episode = []
def Episode_end(self):
last = True
while len(self.episode) > 0:
a = self.episode[-1]['after']
b = self.episode[-1]['before']
R = self.episode[-1]['reward']
S_, S = self.getV(a), self.getV(b)
tmp = Board()
for i in range(8):
tmp.copyBoard(b)
tmp.morphBoard(i)
if last == False:
self.updateNet(tmp, self.alpha*(R + S_ - S))
else:
self.updateNet(tmp, self.alpha*(0 - S))
last = False
del self.episode[-1]
def step(self, prev):
#action = random.randint(0, 3)
#reward = prev.move(action)
#return action, reward
maxV = float(-1e9)
maxOP = -1
tmp = Board()
for op in range(4):
tmp.copyBoard(prev)
r = tmp.move(op)
if r != -1:
v = self.getV(tmp)
if v+r >= maxV:
maxV = v+r
maxOP = op
if maxOP != -1:
r = prev.move(maxOP)
tmp.copyBoard(prev)
state = {
'before': tmp,
'after': tmp,
'reward': r,
'action': maxOP
}
if len(self.episode) > 0:
self.episode[-1]['after'] = tmp
self.episode.append(state)
return maxOP, r
else:
return -1, -1
if __name__ == "__main__":
AI = Agent()
EPISODE = 1001
for e in range(EPISODE):
B = Board()
B.initialize()
while True:
act, r = AI.step(B)
if B.end_game():
break
B.GenRandTile(r)
if B.end_game():
break
#B.printBoard()
if e % 100 == 0:
print("#Episode: {episode}".format(episode = e))
B.printBoard()