def genstate(self):
chg = 1
state0 = [None] * 9
stateV = STATE(state0)
self.pi[tuple(state0)] = stateV
while chg > 0:
n0 = len(self.pi)
keys = self.pi.keys()
for s in keys:
v = self.pi[s]
sgn = getturn(s)
n = len(v.Actions)
for act, v1 in v.Actions.iteritems():
state1 = init.newstate(s, act, sgn)
v1.p = 1. / n
v1.ret = init.evalS(state1)
v1.r = Reward(s, state1)
v1.v = v1.r
if v1.ret is None:
key = tuple(state1)
if key not in self.pi:
stateV = STATE(key)
self.pi[key] = stateV
v1.nextstate = self.pi[key]
n1 = len(self.pi)
chg = n1 - n0
print n0, n1
def rSA(self, state, act, sgn):
sgn1 = 'X' if self.sgn == 'O' else 'O'
state1 = newstate(state, act, sgn)
r0 = self.reward(state, state1)
v = r0
ret = init.evalS(state1)
if ret is None:
r1 = 0
for act1, p1 in self.pi[state1].iteritems():
state2 = newstate(state1, act1, sgn1)
p = self.pi[state1][act1]
V1 = self.value(state2)
r1 += p * V1[0]
v += self.gamma * r1
return v
def score(state):
ret = init.evalS(state)
cnt = 0.
if ret is None:
pt = [1, 1, 1,
1, 1, 1,
1, 1, 1]
for i, s in enumerate(state):
if s:
j = 1 if s == 'O' else -1
cnt += pt[i] * j
elif ret == 'O':
cnt = 10.
elif ret == 'X':
cnt = -10.
return cnt