def __init__(self, sheet):
wetlab = wet_lab.WetLab(num_patients=sheet.num_patients)
sampler = smc.SmcSampler(num_particles=sheet.num_patients * 150,
resample_at_each_iteration=True)
policy_name = sheet.params['policy']
if policy_name == 'G-MIMAX':
policy = group_policy.MimaxPolicy(forward_iterations=2,
backward_iterations=1)
elif policy_name == 'InformativeDorfman':
policy = group_policy.InformativeDorfmanPolicy(cut_off_high=0.95,
cut_off_low=0.001)
elif policy_name == 'Random':
policy = group_policy.MezardPolicy()
else:
raise ValueError(f'Unsupported policy {policy_name}')
super().__init__(
wetlab=wetlab,
sampler=sampler,
policy=policy,
num_simulations=1,
num_tests_per_cycle=sheet.params['tests per cycle'],
max_test_cycles=sheet.params['cycles'],
max_group_size=sheet.params['max group size'],
prior_specificity=sheet.params['specificity'],
prior_sensitivity=sheet.params['sensitivity'],
prior_infection_rate=sheet.priors)
self.sheet = sheet
def __init__(self,
workdir = None,
wetlab=wet_lab.WetLab(),
policy=group_policy.Dorfman(),
sampler=smc.SmcSampler(),
cheap_sampler=lbp.LbpSampler(),
num_simulations = 1,
num_tests_per_cycle = 10,
max_test_cycles = 5,
max_group_size = 8,
prior_specificity = 0.97,
prior_sensitivity = 0.85,
prior_infection_rate = 0.05,
metrics_cls=metrics.Metrics,
export_metrics_every = 5):
self._wetlab = wetlab
self._policy = policy
self._samplers = [cheap_sampler, sampler]
self._max_test_cycles = max_test_cycles
self._num_simulations = num_simulations
self.state = state.State(self._wetlab.num_patients,
num_tests_per_cycle,
max_group_size,
prior_infection_rate,
prior_specificity,
prior_sensitivity)
self.metrics = metrics_cls(
workdir,
self._num_simulations,
self._max_test_cycles,
self._wetlab.num_patients,
self.state.num_tests_per_cycle)
self._export_every = export_metrics_every
def test_selector_with_particles(self, selector):
sampler = sequential_monte_carlo.SmcSampler()
rngs = jax.random.split(self.rng, 2)
sampler.produce_sample(rngs[0], self.state)
self.state.update_particles(sampler)
self.assertEqual(np.size(self.state.groups_to_test), 0)
selector(rngs[1], self.state)
self.assertGreater(np.size(self.state.groups_to_test), 0)
def test_sequential_monte_carlo(self, kernel):
num_particles = 100
sampler = sequential_monte_carlo.SmcSampler(
num_particles=num_particles, kernel=kernel)
self.assertIsNone(sampler.particles)
self.assertIsNone(sampler.particle_weights)
sampler.produce_sample(self.rng, self.state)
self.assertIsNotNone(sampler.particles)
self.assertIsNotNone(sampler.particle_weights)
self.assertEqual(sampler.particles.shape,
(num_particles, self.state.num_patients))
def __init__(self, sheet, num_particles=10000, policy=None):
wetlab = wet_lab.WetLab(num_patients=sheet.num_patients)
sampler = smc.SmcSampler(num_particles=num_particles,
resample_at_each_iteration=True)
if policy is None:
policy = group_policy.MimaxPolicy()
super().__init__(wetlab=wetlab,
sampler=sampler,
policy=policy,
num_simulations=1,
num_tests_per_cycle=sheet.params['tests per cycle'],
max_test_cycles=sheet.params['cycles'],
max_group_size=sheet.params['max group size'],
prior_specificity=sheet.params['specificity'],
prior_sensitivity=sheet.params['sensitivity'],
prior_infection_rate=sheet.priors)
self.sheet = sheet
def __init__(self,
workdir=None,
wetlab=wet_lab.WetLab(),
policy=group_policy.Dorfman(),
sampler=smc.SmcSampler(),
cheap_sampler=lbp.LbpSampler(),
num_simulations=1,
num_tests_per_cycle=10,
max_test_cycles=5,
max_group_size=8,
prior_specificity=0.97,
prior_sensitivity=0.85,
prior_infection_rate=0.05,
metrics_cls=metrics.Metrics,
export_metrics_every=5):
"""Initializes simulation.
Args:
workdir: where results will be stored.
wetlab: WetLab objet tasked with producing test results given groups.
policy: group testing policy, a sequence of algorithms tasked with
choosing groups to test. Can be adaptive to test environment (spec/sens)
of tests. Can be adaptive to previously tested groups. Can leverage
samplers to build information on what are the most likely disease status
among patients.
sampler: sampler that produces a posterior approximation. Instantiated by
default to SmcSampler that resamples at each iteration to fix cases
where the LBP sampler does not quite work the way it should.
cheap_sampler: LBP object to yield cheap approximation of marginal.
num_simulations: number of simulations run consecutively. Here randomness
can come from the group testing policy (if it uses randomness), as well
as diseased label, if WetLab.freeze_diseased is False.
num_tests_per_cycle: number of tests a testing machine can carry out in
the next testing cycle.
max_test_cycles: number of cycles in total that should be considered.
max_group_size: maximal size of groups, how many individuals can be pooled
prior_specificity: best guess one has prior to simulation of the testing
device's specificity per group size (or for all sizes, if singleton).
prior_sensitivity: best guess one has prior to simulation of the testing
device's sensitivity per group size (or for all sizes, if singleton).
prior_infection_rate: best guess of prior probability for patient to be
infected (same for all patients, if singleton)
metrics_cls: class of metrics object used to store results.
export_metrics_every: frequency of exports to file when carrying our
num_simulations results.
"""
self._wetlab = wetlab
self._policy = policy
self._samplers = [cheap_sampler, sampler]
self._max_test_cycles = max_test_cycles
self._num_simulations = num_simulations
self.state = state.State(self._wetlab.num_patients,
num_tests_per_cycle, max_group_size,
prior_infection_rate, prior_specificity,
prior_sensitivity)
self.metrics = metrics_cls(workdir, self._num_simulations,
self._max_test_cycles,
self._wetlab.num_patients,
self.state.num_tests_per_cycle)
self._export_every = export_metrics_every