-
Notifications
You must be signed in to change notification settings - Fork 0
/
rnnac.py
74 lines (65 loc) · 2.02 KB
/
rnnac.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
# coding: UTF-8
import numpy
import random
def r_11():
return random.random()*2-1
class rnn:
def __init__(self, io, h):
self.i_hist = []
self.h_hist = []
self.o_hist = []
self.io = io
self.h = h
self.lh = numpy.array([0. for i in xrange(h)])
self.lhw = numpy.array([[r_11() for i in xrange(h+1)] for k in xrange(h)])
self.hw = numpy.array([[r_11() for i in xrange(io+1)] for k in xrange(h)])
self.ow = numpy.array([[r_11() for i in xrange(h+1)] for k in xrange(io)])
def reset(self):
self.lh = numpy.array([0 for i in xrange(self.h)])
def act(self, x):
return 1/(1+numpy.exp(-x))
def actp(self, y):
return y-numpy.square(y)
def calc(self, i):
th = self.act(sum((numpy.append(i, 1)*self.hw).transpose()))#+sum((self.lh*self.lhw).transpose()))
self.lh = th
to = self.act(sum((numpy.append(th, 1)*self.ow).transpose()))
self.i_hist.append(i)
self.h_hist.append(th)
self.o_hist.append(to)
return to
def update_w(self):
do = []
l = len(self.i_hist)
for i in range(l):
do.append([])
do[-1].append((self.o_hist[i]-self.i_hist[i])*self.actp(self.o_hist[i]))
do[-1].append(numpy.dot(self.hw[:,:-1], do[-1][0])*self.actp(self.h_hist[i]))
self.i_hist = []
omean = sum(self.o_hist)/l
self.o_hist = []
hmean = sum(self.h_hist)/l
self.h_hist = []
sow = numpy.array([0. for i in xrange(self.io)])
shw = numpy.array([0. for i in xrange(self.h)])
for d in do:
sow += d[0]
shw += d[1]
sow /= l
shw /= l
deltawo = sow*omean
deltawh = shw*hmean
self.hw[:,:-1] = self.hw[:,:-1]-0.1*((self.hw[:,:-1].transpose()*deltawh).transpose())-0.00001*self.hw[:,:-1]
self.hw[:,-1] = self.hw[:,-1]-0.1*deltawh
self.ow[:,:-1] = self.ow[:,:-1]-0.1*((self.ow[:,:-1].transpose()*deltawo).transpose())-0.00001*self.ow[:,:-1]
self.ow[:,-1] = self.ow[:,-1]-0.1*deltawo
r = rnn(5, 80)
print r.calc([1, 1, 1, 0, 0])
for i in range(10000):
for k in xrange(10):
r.calc([1, 1, 1, 0, 0])
r.update_w()
#if i%10 == 0:
# raw_input()
print r.calc([1, 1, 1, 0, 0])
print r.calc([1, 1, 1, 0, 0])