def __init__(self, weights, starts, length, nK, nSigma, d, folderPath):
self.weights = weights
self.generator = CmCmGenerator(length, starts[0], starts[1])
self.nSigma = nSigma
self.nk = nK
self.dk = self.dSigma = d
self.pathToFolder = folderPath
class Orchestrator():
def __init__(self, weights, starts, length, nK, nSigma, d, folderPath):
self.weights = weights
self.generator = CmCmGenerator(length, starts[0], starts[1])
self.nSigma = nSigma
self.nk = nK
self.dk = self.dSigma = d
self.pathToFolder = folderPath
def pi_from_sigma_k(self, parameter):
# Создадим массив коэффициентов
self.generator.coefficients = [parameter] + self.weights
mPI_s_k = mEself_s_k = mEjoin_s_k = empty((self.nk, self.nSigma))
# Создаём папку
if not os.path.exists(self.pathToFolder):
os.makedirs(self.pathToFolder)
for cf in xrange(self.nk):
# Сгенерируем тестовые сигналы
(radX, radY) = self.generator.generate(cf * self.dk)
# Зашумим сигналы
for sigma in xrange(self.nSigma):
startTime = time()
# Усредним занчения PI по сотне точек
mPI_average = mEself_average = mEjoin_average = []
for ll in xrange(100):
# Добавляем шум:
ssigma = (sigma + 1) * self.dSigma
radXNoise = radX + random.normal(0, ssigma, len(radX))
radYNoise = radY + random.normal(0, ssigma, len(radY))
# Рассчитаем ошибки аппроксимации и PI:
fitS = CommonUtils.fit_self(radXNoise, 3)[1]/(len(radX)-1)
fitJ = CommonUtils.fit_join(radXNoise, radYNoise, 3)[1]/(len(radX)-1)
if len(fitJ) > 0 and len(fitS) > 0:
mPI_average.append(1 - fitJ[0]/fitS[0])
mEself_average.append(fitS[0])
mEjoin_average.append(fitJ[0])
# Подсчитаем среднее значение PI для данного сигма:
if mPI_average:
mPI_s_k[cf, sigma] = mean(require(mPI_average))
else:
mPI_s_k[cf, sigma] = 0
if mEself_average:
mEself_s_k[cf, sigma] = mean(require(mEself_average))
else:
mEself_s_k[cf, sigma] = 0
if mEjoin_average:
mEjoin_s_k[cf, sigma] = mean(require(mEjoin_average))
else:
mEjoin_s_k[cf, sigma] = 0
print "sigma=%4.2f, k=%4.2f, time=%6.3f" % (sigma, cf*self.dk, time()-startTime)
# Сохраним полученный массив в файл:
f = open(os.path.join(self.pathToFolder, "PI-from-sigma"+"%4.2f" % (cf*self.dk)+".txt"), 'w')
for ii in xrange(self.nSigma):
f.write("%4.2f" % ((ii+1)*self.dSigma) + ' ' + "%14.10f" % mPI_s_k[cf, ii] + '\n')
f.close()
# Сохраним полученный массив в файл:
f2 = open(os.path.join(self.pathToFolder, "Eself-from-sigma"+"%4.2f" % (cf*self.dk)+".txt"), 'w')
for ii in xrange(self.nSigma):
f2.write("%4.2f" % ((ii+1)*self.dSigma) + ' ' + "%14.10f" % mEself_s_k[cf, ii] + '\n')
f2.close()
# Сохраним полученный массив в файл:
f3 = open(os.path.join(self.pathToFolder, "Ejoin-from-sigma"+"%4.2f" % (cf*self.dk)+".txt"), 'w')
for ii in xrange(self.nSigma):
f3.write("%4.2f" % ((ii+1)*self.dSigma) + ' ' + "%14.10f" % mEjoin_s_k[cf, ii] + '\n')
f3.close()
# Записываем чистый ряд Y:
with open(os.path.join(self.pathToFolder, "radY_"+"%4.2f" % (cf)+".txt"), 'w') as fy:
for ii in xrange(len(radY)-1):
fy.write("%12.8f %12.8f\n" % (radY[ii], radY[ii+1]))
# Записываем чистый ряд X при не нулевой силе связи:
with open(os.path.join(self.pathToFolder, "radX_"+"%4.2f" % (cf)+".txt"), 'w') as fx:
for ii in xrange(len(radX)-1):
fx.write("%12.8f %12.8f\n" % (radX[ii], radX[ii+1]))
sk = argmin(mPI_s_k[:, :], axis=1) * self.dSigma
fsk = open(os.path.join(self.pathToFolder, "sk.txt"), 'w')
for s in xrange(self.nk):
fsk.write("%12.8f %12.8f" % (self.dk*s, sk[s])+"\n")
fsk.close()
return True
