def netdem(source, target, ndiff, npack = 10, nlearn = 500, epoch = 100):
diffx = []
# for i in xrange(len(source)):
# val = 0
# for j in xrange(ndiff):
# if(i + j < len(source)):
# val += source[i+j]
# diffx += [(val/ndiff)]
# sample = diffx
sample = source
'''x = []
for i in xrange(len(sample)):
frame = []
for j in xrange(npack):
if (i-j) < 0:
frame += [0]
else:
frame += [sample[i-j]]
x += [frame]'''
x = [sample[i-npack:i] for i in xrange(npack, len(sample)+1, npack)]
test = [[target[i]] for i in xrange(len(target))]
print 'LEN source = ', len(source)
print 'LEN X = ', len(x)
print 'LEN test = ', len(test)
pnet = per_net([npack, 4, 1], elman = 1)
epoch = pnet.per_train(x[:nlearn], test[:nlearn], epoch, 0.001, 0.01)
return pnet, x
from per_net.per_net import *
import sys
x = [[0, 0, 0, 0],
[0, 0, 0, 1],
[0, 0, 1, 1],
[0, 1, 1, 1],
[1, 1, 1, 1]]
t = [[0, 0, 0, 0, 0, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 1, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 0, 0, 0]]
a_in = float(sys.argv[2])
n_in = float(sys.argv[3])
c_in = int(sys.argv[1])
p = per_net([4, 20, 9])
epoch = p.per_train(x, t, c_in, n = n_in, alph = a_in)
test_x = [[1, 1, 1, 0]]
if(epoch < 999):
for i in xrange(len(test_x)):
print (p.sim_net(test_x[i]))[p.count_layer -1]
Fd = 2400 # символьная скорость
Fc = 1800 # несущая
FsFd = 4 # количество отсчетов на один символ
Fs = Fd * FsFd # частота дискретизации
t = [i/Fs for i in range(N*FsFd)] # дискретное время
x = [2*math.pi*Fc*i for i in t]
x = [[x_tmp[i]/10.0] for i in xrange(len(x_tmp))]
t = [[math.sin(x[i][0])] for i in xrange(len(x_tmp))]
print x
print t
a_in = float(sys.argv[2])
n_in = float(sys.argv[3])
c_in = int(sys.argv[1])
p = per_net([1, 5, 2, 1])
epoch = p.per_train(x, t, c_in, n = n_in, alph = a_in)
x_tmp = xrange(-10, 10, 5)
test_x = [[x_tmp[i]/10.0] for i in xrange(len(x_tmp))]
print test_x
if(epoch < 999):
for i in xrange(len(test_x)):
print (p.sim_net(test_x[i]))[p.count_layer -1]
