def get_init_fun(args, stan_data, force_fresh=False):
if args.init and not force_fresh:
try:
init_path = Path(args.fits_path) / args.init
model_path = Path(args.models_path) / args.model_name
result = ncs.last_sample_as_dict(init_path, model_path)
except Exception:
print("Couldn't use last sample from previous fit to initialize")
return init_fun(force_fresh=True)
else:
print("Using last sample from previous fit to initialize")
else:
print("Using default values to initialize fit")
result = {
'sigmau': exponential(1.)
# 'beta': normal(1.,.5),
# 'sigr': exponential(.5),
# 'sigd': exponential(.2),
# 'sigmau': exponential(1.),
# 'sigc': exponential(1.),
# 'alpha': exponential(1.)
}
def init_fun():
return result
return init_fun
def get_init_funX(args, stan_data, force_fresh=False):
if args.init and not force_fresh:
try:
init_path = Path(args.fits_path) / args.init
model_path = Path(args.models_path) / args.model_name
result = ncs.last_sample_as_dict(init_path, model_path)
except Exception:
print("Couldn't use last sample from previous fit to initialize")
return init_fun(force_fresh=True)
else:
print("Using last sample from previous fit to initialize")
else:
print("Using default values to initialize fit")
result = {
'f1': gamma(2., 10.),
'f2': gamma(40., 1 / 100.),
'sigmar': gamma(20, 1 / 120.),
'sigmad': gamma(20, 1 / 120),
'sigmau': gamma(2., 1 / 20.),
'q': exponential(.1),
'mbase': gamma(2., .1 / 2.),
# 'mlocation': lognormal(np.log(stan_data['tm']), 1.),
'mlocation': normal(stan_data['tm'], 4.),
'extra_std': exponential(.5),
'extra_std_R': exponential(.5),
'extra_std_D': exponential(.5),
'cbase': gamma(1., 1.),
# 'clocation': lognormal(np.log(20.), 1.),
'clocation': normal(50., 1.),
'ctransition': normal(10., 1.),
# 'n_pop': lognormal(np.log(1e5), 1.),
'n_pop': normal(1e6, 1e4),
'sigmar1': gamma(2., .01),
'sigmad1': gamma(2., .01),
'trelax': normal(50., 5.)
}
def init_fun():
return result
return init_fun
def init_fun(force_fresh=False):
if args.init and not force_fresh:
try:
init_path = Path(args.fits_path) / args.init
model_path = Path(args.models_path) / args.model_name
result = ncs.last_sample_as_dict(init_path, model_path)
except Exception:
print("Couldn't use last sample from previous fit to initialize")
return init_fun(force_fresh=True)
else:
print("Using last sample from previous fit to initialize")
else:
print("Using default values to initialize fit")
from numpy.random import gamma, exponential, lognormal
result = {'f1': gamma(2.4,10.),
'f2': gamma(175,1/420.),
'sigmar': gamma(16.,1/120.),
'sigmad': gamma(19.,1/1200.),
'sigmau': gamma(2.,1/23.),
'q': exponential(.1),
'mbase': gamma(10.,1/30.),
}
return result
def get_init_funV(args, stan_data, force_fresh=False):
if args.init and not force_fresh:
try:
init_path = Path(args.fits_path) / args.init
model_path = Path(args.models_path) / args.model_name
result = ncs.last_sample_as_dict(init_path, model_path)
except Exception:
print("Couldn't use last sample from previous fit to initialize")
return init_fun(force_fresh=True)
else:
print("Using last sample from previous fit to initialize")
else:
print("Using default values to initialize fit")
result = {
'f1': 1.75,
'f2': .25,
'sigmaVS': exponential(1 / 10.),
'sigmaSR': exponential(1 / 400.),
'sigmaSRI': exponential(1 / 400.),
'sigmaIV': exponential(1 / 100.),
'sigmaRI': .01,
'sigmaDI': .03,
'sigmaRC': .08,
'sigmaDC': .003,
'sigmaRH1': 1.,
'sigmaRH2': .1,
'sigmaRV': exponential(1 / 10.),
'sigmaH1C': .01,
'sigmaH1V': exponential(1 / 10.),
'sigmaH2H1': .4,
'sigmaDH1': .003,
'sigmaDH2': .001,
'sigmaM': exponential(1 / 10.),
'mbase': exponential(1 / 10.),
'q': exponential(.5),
'n_pop': normal(4e6, 1e4)
}
# result = {'f1': exponential(1.),
# 'f2': exponential(1.),
# 'sigmaVS': exponential(1/10.),
# 'sigmaSR': exponential(1/400.),
# 'sigmaSRI': exponential(1/400.),
# 'sigmaIV': exponential(1/100.),
# 'sigmaRI': exponential(1/10.),
# 'sigmaDI': exponential(1/30.),
# 'sigmaRC': exponential(1/10.),
# 'sigmaDC': exponential(1/10.),
# 'sigmaRH1': exponential(1/10.),
# 'sigmaRH2': exponential(1/10.),
# 'sigmaRV': exponential(1/10.),
# 'sigmaH1C': exponential(1/10.),
# 'sigmaH1V': exponential(1/10.),
# 'sigmaH2H1': exponential(1/10.),
# 'sigmaRH1': exponential(1/10.),
# 'sigmaDH1': exponential(1/10.),
# 'sigmaDH2': exponential(1/10.),
# 'sigmaM': exponential(1/10.),
# 'mbase': exponential(1/10.),
# 'q': exponential(.1),
# 'n_pop': normal(1e6, 1e4)
# }
def init_fun():
return result
return init_fun