def __init__(self, mec, recs):
self.mec = mec
self.recs = recs
self.LL = []
kwargs = {'nmax': 2, 'xtol': 1e-12, 'rtol': 1e-12, 'itermax': 100,
'lower_bound': -1e6, 'upper_bound': 0}
print (self.recs[0].bursts.intervals())
for i in range(len(self.recs)):
self.LL.append(Log10Likelihood(self.recs[i].bursts.intervals(), self.mec.kA,
self.recs[i].tres, self.recs[i].tcrit, **kwargs))
self.report = FitReportHTML()
self.report.dataset(self.recs)
start_lik = self.dcprogslik(np.log(self.mec.theta()))
print ("\nStarting likelihood = {0:.6f}".format(-start_lik))
start_lik_all = self.dcprogslik_all(np.log(self.mec.theta()))
for i in range(len(start_lik_all)):
print ("Set #{0:d}: likelihood = {1:.6f}".format(i+1, -start_lik_all[i]))
print ('\n\n')
self.report.rates(self.mec, start_lik)
self.iternum = 0
class FittingSession():
"""
"""
def __init__(self, mec, recs):
self.mec = mec
self.recs = recs
self.LL = []
kwargs = {'nmax': 2, 'xtol': 1e-12, 'rtol': 1e-12, 'itermax': 100,
'lower_bound': -1e6, 'upper_bound': 0}
print (self.recs[0].bursts.intervals())
for i in range(len(self.recs)):
self.LL.append(Log10Likelihood(self.recs[i].bursts.intervals(), self.mec.kA,
self.recs[i].tres, self.recs[i].tcrit, **kwargs))
self.report = FitReportHTML()
self.report.dataset(self.recs)
start_lik = self.dcprogslik(np.log(self.mec.theta()))
print ("\nStarting likelihood = {0:.6f}".format(-start_lik))
start_lik_all = self.dcprogslik_all(np.log(self.mec.theta()))
for i in range(len(start_lik_all)):
print ("Set #{0:d}: likelihood = {1:.6f}".format(i+1, -start_lik_all[i]))
print ('\n\n')
self.report.rates(self.mec, start_lik)
self.iternum = 0
def likelihood(self):
for i in range(len(self.recs)):
self.LL.append(Log10Likelihood(self.recs[i].bursts, self.mec.kA,
self.recs[i].tres, self.recs[i].tcrit, **self.kwargs))
def run(self):
# FITTING BIT
start = time.clock()
result = minimize(self.dcprogslik, np.log(self.mec.theta()),
method='Nelder-Mead', callback=self.printiter,
options={'xtol':1e-4, 'ftol':1e-4, 'maxiter': 5000, 'maxfev': 10000,
'disp': True})
end = time.clock() - start
# FITTING BIT
print ("\n\nFitting with DCPROGS likelihood finished: %4d/%02d/%02d %02d:%02d:%02d"
%time.localtime()[0:6])
print ('\n Time in simplex=', end)
print ('\n Final log-likelihood = {0:.6f}'.format(-result.fun))
print ('\n Number of iterations = {0:d}'.format(result.nit))
self.mec.theta_unsqueeze(np.exp(result.x))
print ("\n Final rate constants:")
self.mec.printout(sys.stdout)
self.report.fit_result(self.mec, start, end, result.fun, result.nit)
self.report.rates(self.mec, result.fun, True)
self.report.distributions(self.mec, self.recs)
self.report.finalise()
def printiter(self, theta):
self.iternum += 1
lik = self.dcprogslik(theta)
if self.iternum % 10 == 0:
print("iteration # {0:d}; log-lik = {1:.6f}".format(self.iternum, -lik))
print(np.exp(theta))
def dcprogslik(self, x, args=None):
self.mec.theta_unsqueeze(np.exp(x))
lik = 0
for i in range(len(self.recs)):
self.mec.set_eff('c', self.recs[i].conc)
lik += -self.LL[i](self.mec.Q) * log(10)
return lik
def dcprogslik_all(self, x, args=None):
"""
Return a list of separate likelihoods for each patch.
"""
self.mec.theta_unsqueeze(np.exp(x))
lik = []
for i in range(len(self.recs)):
self.mec.set_eff('c', self.recs[i].conc)
lik.append(-self.LL[i](self.mec.Q) * log(10))
return lik