def run_experiment(path: Path, custom_settings: dict):
# 1. Initialize the experiment
exp = Experiment()
exp.load_config(config_file=path / 'config.json',
custom_settings=custom_settings)
if exp.config['model']['n_areas'] == "TBD":
exp.config['model']['n_areas'] = 2
# 2. Simulate contact areas
sim = Simulation(experiment=exp)
sim.run_simulation()
# 3. Define MCMC
mc = MCMC(data=sim, experiment=exp)
# 4. Warm-up sampler and sample from posterior
mc.warm_up()
mc.sample()
# 5. Evaluate ground truth
mc.eval_ground_truth()
# 6. Log sampling statistics and save samples to file
mc.save_samples(run=0)
def run_experiment(experiment, data, run, initial_sample=None):
mcmc = MCMC(data=data, experiment=experiment)
mcmc.log_setup()
# Sample
mcmc.sample(initial_sample=initial_sample)
# Save samples to file
mcmc.log_statistics()
mcmc.save_samples(run=run)
# Use the last sample as the new initial sample
return mcmc.samples['last_sample']
# 1. Initialize the experiment
exp = Experiment()
exp.load_config(
config_file='experiments/simulation/sim_exp2/config.json',
custom_settings={'model': {
'INHERITANCE': IN
}})
exp.log_experiment()
# 2. Simulate contact areas
sim = Simulation(experiment=exp)
sim.run_simulation()
sim.log_simulation()
# 3. Define MCMC
mc = MCMC(data=sim, experiment=exp)
mc.log_setup()
# Rerun experiment to check for consistency
for run in range(exp.config['mcmc']['N_RUNS']):
# 4. Sample from posterior
mc.warm_up()
mc.sample()
# 5. Evaluate ground truth
mc.eval_ground_truth()
# 6. Log sampling statistics and save samples to file
mc.log_statistics()
mc.save_samples(run=run)
dat = Data(experiment=exp)
# Features
dat.load_features()
# Counts for priors
dat.load_universal_counts()
dat.load_inheritance_counts()
# Log
dat.log_loading()
NUMBER_AREAS = range(1, 8)
# Rerun experiment to check for consistency
for run in range(exp.config['mcmc']['N_RUNS']):
for N in NUMBER_AREAS:
# Update config information according to the current setup
exp.config['model']['N_AREAS'] = N
# MCMC
mc = MCMC(data=dat, experiment=exp)
mc.log_setup()
# Sample
mc.warm_up()
mc.sample()
# Save samples to file
mc.log_statistics()
mc.save_samples(run=run)
def run_experiment(experiment, data, run):
mcmc = MCMC(data=data, experiment=experiment)
mcmc.log_setup()
# Warm-up
mcmc.warm_up()
# Sample from posterior
mcmc.sample()
# Save samples to file
mcmc.log_statistics()
mcmc.save_samples(run=run)
# Use the last sample as the new initial sample
return mcmc.samples['last_sample']