def loglikelyhood_SI_relic(x):
dest = DiffusionEstimator.default(echo=True)
if x[0] < 1e-12 or x[1] < 1e-12:
return 1e20
ll = dest.loglikelyhood(x[0], 0., 0., x[1])
#print(str(-ll) + ' ' + str(x))
return -ll
def llmax_for_diffusion_noconfirm_ensemble(theta0,
decay0,
relic0,
observe_time=-1,
disp=True,
new_lib=False):
dest = DiffusionEstimator.default(echo=disp)
x0 = np.array([theta0, decay0, relic0])
if new_lib:
res = minimize(loglikelyhood_noconfirm_ensemble,
x0,
args=(observe_time, ),
method='nelder-mead',
options={
'xtol': 1e-11,
'disp': disp
})
else:
res = minimize(loglikelyhood_noconfirm_ensemble,
x0,
args=(observe_time, ),
method='nelder-mead',
options={
'xtol': 1e-11,
'disp': disp
})
return {'theta': res.x[0], 'decay': res.x[1], 'relic': res.x[2]}
def loglikelyhood_noconfirm_ensemble(x, observe_time=-1):
dest = DiffusionEstimator.default(echo=True)
if (x[0] < 1e-12 and x[2] < 1e-12) or x[1] < 0:
return 1e20
ll = dest.loglikelyhood_ensemble(x[0], 0., x[1], x[2], observe_time)
#print(str(-ll) + ' ' + str(x))
return -ll
def loglikelyhood(x):
dest = DiffusionEstimator.default(echo=True)
if x[0] < 1e-12 or x[3] < 1e-12 or x[2] < 0 or x[1] < -1. or x[1] > 1.:
return 1e20
ll = dest.loglikelyhood(x[0], x[1], x[2], x[3])
print(str(-ll) + ' ' + str(x))
return -ll
def llfunc_for_diffusion(thetas, confirms, decays, relics):
dest = DiffusionEstimator.default(echo=True)
ps = np.zeros((len(thetas), len(confirms), len(decays), len(relics)))
for i in range(len(thetas)):
for j in range(len(confirms)):
for k in range(len(decays)):
for l in range(len(relics)):
ps[i][j][k][l] = dest.loglikelyhood(
thetas[i], confirms[j], decays[k], relics[l])
return ps
def loglikelyhood_noconfirm(x, observe_time=-1.):
dest = DiffusionEstimator.default(echo=True)
if x[0] < 1e-12 or x[2] < 1e-12 or x[1] < 0:
return 1e20
start_time = time.time()
ll = dest.loglikelyhood(x[0], 0., x[1], x[2], observe_time)
end_time = time.time()
print(
str(-ll) + ' ' + str(x) + ' ex. time: ' +
str(end_time - start_time))
return -ll
def llmax_for_diffusion_noconfirm_by_node_theta(node_id,
theta,
decay,
relic,
bounds,
observe_time=0,
disp=True):
dest = DiffusionEstimator.default(echo=True)
res = minimize_scalar(loglikelyhood_by_node_theta,
bounds=bounds,
method='bounded',
args=(node_id, theta, decay, relic, observe_time))
return res.x
def llmax_for_diffusion_noconfirm_multiple_cases(thetas0,
decay0,
relics0,
disp=True):
dest = DiffusionEstimator.default(echo=disp)
sx = thetas0 + relics0 + [decay0]
x0 = np.array(sx)
res = minimize(loglikelyhood_noconfirm_newlib_multiple_cases,
x0,
method='nelder-mead',
options={
'xtol': 1e-11,
'disp': disp
})
return {'theta': res.x[0], 'decay': res.x[1], 'relic': res.x[2]}
def llmax_for_diffusion(theta0, confirm0, decay0, relic0):
dest = DiffusionEstimator.default(echo=True)
x0 = np.array([theta0, confirm0, decay0, relic0])
res = minimize(loglikelyhood,
x0,
method='nelder-mead',
options={
'xtol': 1e-11,
'disp': True
})
return {
'theta': res.x[0],
'confirm': res.x[1],
'decay': res.x[2],
'relic': res.x[3]
}
def llmax_for_diffusion_SI_relic(theta0, relic0, delta=.0, new_lib=False):
dest = DiffusionEstimator.default(echo=True)
x0 = np.array([theta0, relic0])
if new_lib:
res = minimize(loglikelyhood_SI_relic_newlib,
x0,
args=(delta, ),
method='nelder-mead',
options={
'xtol': 1e-11,
'disp': True
})
else:
res = minimize(loglikelyhood_SI_relic,
x0,
method='nelder-mead',
options={
'xtol': 1e-11,
'disp': True
})
return {'theta': res.x[0], 'relic': res.x[1]}
def llmax_for_diffusion_ICM(theta0, relic0=.0):
dest = DiffusionEstimator.default(echo=True)
if relic0 == .0:
x0 = np.array([theta0])
res = minimize(loglikelyhood_ICM_newlib,
x0,
method='nelder-mead',
options={
'xtol': 1e-11,
'disp': True
})
return {'theta': res.x[0]}
else:
x0 = np.array([theta0, relic0])
res = minimize(loglikelyhood_ICM_newlib,
x0,
method='nelder-mead',
options={
'xtol': 1e-11,
'disp': True
})
return {'theta': res.x[0], 'relic': res.x[1]}
def llmax_for_diffusion_SI_decay(theta0, delta0=.0, fixed_delta=False):
dest = DiffusionEstimator.default(echo=True)
if fixed_delta == True:
x0 = np.array([theta0])
res = minimize(loglikelyhood_SI_decay_newlib,
x0,
method='nelder-mead',
options={
'xtol': 1e-11,
'disp': True
},
args=(delta0, ))
return {'theta': res.x[0]}
else:
x0 = np.array([theta0, delta0])
res = minimize(loglikelyhood_SI_decay_newlib,
x0,
method='nelder-mead',
options={
'xtol': 1e-11,
'disp': True
})
return {'theta': res.x[0], 'decay': res.x[1]}
def set_diffusion_data_ensemble(netwf: str, outcomes, counters, echo=True):
dest = DiffusionEstimator.default(echo=echo)
dest.load_network_inf_ensemble(netwf, counters=counters, outcomes=outcomes)
def dlldthetas_noconfirm(theta, decay, relic, ids, observe_time=.0):
dest = DiffusionEstimator.default(echo=True)
dll = dest.thetas_derivatives(theta, .0, decay, relic, ids, observe_time)
#print(str(dll) + ' ' + str(x))
return dll
def set_diffusion_data(netwf: str, outcome, counters, echo=True):
dest = DiffusionEstimator.default(echo=echo)
dest.load_network(netwf, counters=counters, outcome=outcome)
def loglikelyhood_by_node_theta(x, node_id, theta, decay, relic, observe_time):
dest = DiffusionEstimator.default(echo=True)
ll = dest.loglikelyhood_by_node_theta(x, node_id, theta, .0, decay, relic,
observe_time)
#print(str(-ll) + ' ' + str(x))
return -ll