# FOR TESTING ONLY:
if mode == 'fast':
cb.set_param('x_Temporary_Larval_Habitat', 0)
cb.set_param('x_Regional_Migration', 0)
return sample
if calib_stage == 1:
sigma_r = 0.2
elif calib_stage >= 2:
sigma_r = 0.05
if calib_stage != 0:
optimtool = OptimTool(params,
samples_per_iteration=samples_per_iteration,
center_repeats=1,
sigma_r=sigma_r)
calib_manager = CalibManager(
name=COMPS_calib_exp_name,
config_builder=mozamb_exp.cb,
map_sample_to_model_input_fn=map_sample_to_model_input,
sites=sites,
next_point=optimtool,
sim_runs_per_param_set=sim_runs_per_param_set,
max_iterations=max_iterations,
plotters=plotters)
run_calib_args = {"calib_manager": calib_manager}
if __name__ == "__main__":
"Vector_Species_Names": [species]
})
# Just for fun, let the numerical derivative baseline scale with the number of dimensions
volume_fraction = 0.01 # desired fraction of N-sphere area to unit cube area for numerical derivative (automatic radius scaling with N)
num_params = len([p for p in params if p['Dynamic']])
#Change radius
r = math.exp(1/float(num_params)*(math.log(volume_fraction) + gammaln(num_params/2.+1) - num_params/2.*math.log(math.pi)))
#r = 0.5
#r *= 0.5
optimtool = OptimTool(params,
mu_r = r, # <-- radius for numerical derivatve. CAREFUL not to go too small with integer parameters
sigma_r = r/10., # <-- stdev of radius
center_repeats=1, # <-- Number of times to replicate the center (current guess). Nice to compare intrinsic to extrinsic noise
samples_per_iteration=1000 # was 32 # <-- Samples per iteration, includes center repeats. Actual number of sims run is this number times number of sites.
)
# cb.add_reports(BaseVectorStatsReport(type='ReportVectorStats', stratify_by_species=1))
add_filtered_report(cb, start=365*throwaway)
calib_manager = CalibManager(name=expname, # <-- Please customize this name
config_builder=cb,
map_sample_to_model_input_fn=map_sample_to_model_input,
sites=sites,
next_point=optimtool,
sim_runs_per_param_set=1, # <-- Replicates
max_iterations=max_iterations, # <-- Iterations
plotters=plotters)
num_params = len([p for p in params if p['Dynamic']])
r = math.exp(1 / float(num_params) *
(math.log(volume_fraction) + gammaln(num_params / 2. + 1) -
num_params / 2. * math.log(math.pi)))
# Check, here's the formula for the volume of a N-sphere
computed_volume_fraction = math.exp(num_params / 2. * math.log(math.pi) -
gammaln(num_params / 2. + 1) +
num_params * math.log(r))
optimtool = OptimTool(
params,
constrain_sample, # <-- Will not be saved in iteration state
#mu_r = r, # <-- Mean percent of parameter range for numerical derivatve. CAREFUL with integer parameters!
mu_r=0.3,
sigma_r=r / 10., # <-- stdev of above
samples_per_iteration=500, # 700 is real size, 10 is testing
center_repeats=10, # 10 is real size, 2 is testing
rsquared_thresh=
0.5 # <-- Linear regression goodness of fit threshold. Above this, regression is used. Below, use best point.
)
calib_manager = CalibManager(
name='Rakai_calibration_run4',
config_builder=config_builder,
map_sample_to_model_input_fn=map_sample_to_model_input,
sites=sites,
next_point=optimtool,
sim_runs_per_param_set=3, # <-- Replicates, none needed for example
max_iterations=100, # limited for example
plotters=plotters)
def create_DTK():
''' Create an OptimTool instance '''
# Overwrite default repr with something more informative
def new_repr(self):
return sc.prepr(self)
def gather_results(self, randseed=None, results=None):
if randseed is not None:
pl.seed(randseed)
if results is None:
results = pl.rand(len(self.data['Results']))
self.data['Results'][-len(results):] = results
return results
OptimTool.__repr__ = new_repr
OptimTool.gather_results = gather_results
params = [
{
'Name': 'Clinical Fever',
'Dynamic': True,
#'MapTo': 'Clinical_Fever_Threshold_High', # <-- DEMO: Custom mapping, see map_sample_to_model_input below
'Guess': 1.75,
'Min': 0.5,
'Max': 2.5
},
{
'Name': 'MSP1 Merozoite',
'Dynamic': False, # <-- NOTE: this parameter is frozen at Guess
'MapTo': 'MSP1_Merozoite_Kill_Fraction',
'Guess': 0.65,
'Min': 0.4,
'Max': 0.7
},
{
'Name': 'Falciparum',
'Dynamic': True,
'MapTo': 'Falciparum_PfEMP1_Variants',
'Guess': 1500,
'Min': 1, # 900 [0]
'Max': 5000 # 1700 [1e5]
},
{
'Name': 'Min Days',
'Dynamic': False, # <-- NOTE: this parameter is frozen at Guess
'MapTo': 'Min_Days_Between_Clinical_Incidents',
'Guess': 25,
'Min': 1,
'Max': 50
},
]
num_params = len([p for p in params if p['Dynamic']])
volume_fraction = 0.01 # desired fraction of N-sphere area to unit cube area for numerical derivative (automatic radius scaling with N)
r = OptimTool.get_r(num_params, volume_fraction)
OT = OptimTool(
params,
mu_r=
r, # <-- radius for numerical derivatve. CAREFUL not to go too small with integer parameters
sigma_r=r / 10., # <-- stdev of radius
center_repeats=
0, # <-- Number of times to replicate the center (current guess). Nice to compare intrinsic to extrinsic noise
samples_per_iteration=
10 # 32 # <-- Samples per iteration, includes center repeats. Actual number of sims run is this number times number of sites.
)
return OT
# Replace the (species S 100) with the correct number coming from our sample
if p["Name"] == "initial_S_A04":
cb.set_species("S-A04", int(value))
# Replaces the (species I 100) with the correct number coming from our sample
elif p["Name"] == "initial_I_A04":
cb.set_species("I-A04", int(value))
# Add this change to our tags
tags[p["Name"]] = value
return tags
optimtool = OptimTool(params,
constrain_sample,
samples_per_iteration=25,
center_repeats=1)
calib_manager = CalibManager(
name='ExampleOptimizationCMS',
config_builder=cb,
map_sample_to_model_input_fn=map_sample_to_model_input,
sites=sites,
next_point=optimtool,
sim_runs_per_param_set=1,
max_iterations=3,
plotters=plotters)
run_calib_args = {"calib_manager": calib_manager}
if __name__ == "__main__":
})
def map_sample_to_model_input(cb, sample):
hab = {
'gambiae': {
'TEMPORARY_RAINFALL': 1.5e8 * sample["Temporary_Habitat"],
'CONSTANT': 5.5e6 * sample["Constant_Habitat"]
}
}
set_larval_habitat(cb, hab)
return sample
optimtool = OptimTool(params, center_repeats=1, samples_per_iteration=25)
calib_manager = CalibManager(
name=exp_name,
config_builder=cb,
map_sample_to_model_input_fn=map_sample_to_model_input,
sites=[PointPrevalenceSite(site)],
next_point=optimtool,
sim_runs_per_param_set=1,
max_iterations=3,
plotters=[OptimToolPlotter()])
add_summary_report(
cb,
description='AnnualAverage') ## add summary report of annual prevalence
run_calib_args = {"calib_manager": calib_manager}
funest_rescale *
sample['b{}_vector_mult'.format(int(bairro_num))],
start_day=0)
# FOR TESTING ONLY:
if mode == 'fast':
cb.set_param('x_Temporary_Larval_Habitat', 0)
cb.set_param('x_Regional_Migration', 0)
return sample
if calib_stage != 0:
optimtool = OptimTool(
params,
samples_per_iteration=samples_per_iteration,
center_repeats=1,
sigma_r=0.05
) #increase radius of hypersphere (default is sigma_r=0.02)
calib_manager = CalibManager(
name=COMPS_calib_exp_name,
config_builder=mozamb_exp.cb,
map_sample_to_model_input_fn=map_sample_to_model_input,
sites=sites,
next_point=optimtool,
sim_runs_per_param_set=sim_runs_per_param_set,
max_iterations=max_iterations,
plotters=plotters)
run_calib_args = {"calib_manager": calib_manager}