def create_model(self, ):
mcmc = BayesianMcmc(self.data_model)
mcmc.hyperparams['alpha'] = 2
mcmc.hyperparams['beta'] = 5
mcmc.mh_params['chain_length'] = 2000
mcmc.data_model.bscc_eval_mode = BeesModel.BSCC_MODE_PFUNCS
return mcmc
def test_hpd(data):
sample, expected_hpd = data
mcmc = BayesianMcmc(None)
hpd = mcmc._compute_hpd(sample, 0.95)
print(hpd)
for i in range(0, len(hpd)):
assert hpd[i] - expected_hpd[i] < 1e-9
def mcmc():
dtmc_filepath = 'data/prism/bee_multiparam_synchronous_3.pm'
bscc_filepath = 'data/prism/bee_multiparam_synchronous_3.txt'
model = BeesModel.from_files(dtmc_filepath, bscc_filepath)
mcmc = BayesianMcmc(model)
assert model.bscc_eval_mode == BeesModel.BSCC_MODE_CHAIN_RUN
return mcmc
def do_experiment(self, ):
self.init()
mcmc = BayesianMcmc(self.data_model)
mcmc.data_model.chainruns_count = 400
mcmc.mh_params['chain_length'] = 200
self.print_log('Start inferencing, MCMC chain len {}'.format(1000))
m = self.syn_data['mult']
# start inferencing
start_time = timeit.default_timer()
mcmc.estimate_p(m)
stop_time = timeit.default_timer()
self.print_log('MCMC estimation, ETA(secs) {}'.format(stop_time - start_time))
print(mcmc.estimated_params['P'])
rmse = sqrt(mean_squared_error(self.p_true, mcmc.estimated_params['P']))
self.print_log('RMSE {}'.format(rmse))
self.print_log(mcmc.summarize())
def do_experiment_rational(model, data_m, mh_chain_length):
logging.info('...')
logging.info(
'Metropolis Hastings, get BSCC by rational function evaluation')
mcmc = BayesianMcmc(model)
mcmc.mh_params['chain_length'] = mh_chain_length
mcmc.hyperparams = {
'alpha': 1.5,
'beta': 1.5,
}
start_time = timeit.default_timer()
mcmc.estimate_p(data_m)
stop_time = timeit.default_timer()
elapsed_time = stop_time - start_time
logging.info('Finished in {} seconds'.format(elapsed_time))
summarize_inference_result(mcmc)
def do_experiment_beta(self, ):
self.print_log('BETA PRIOR')
config.models['use_uniform_prior'] = False
self.data_model.bscc_eval_mode = BeesModel.BSCC_MODE_PFUNCS
mcmc = BayesianMcmc(self.data_model)
mcmc_chainlen = 500
mcmc.mh_params['chain_length'] = mcmc_chainlen
mcmc.hyperparams = {'alpha': 2, 'beta': 5}
self.print_log(
'Start inferencing, MCMC chain len {}'.format(mcmc_chainlen))
m = self.syn_data['mult']
start_time = timeit.default_timer()
mcmc.estimate_p(m)
stop_time = timeit.default_timer()
self.print_log('MCMC estimation, ETA(secs) {}'.format(stop_time -
start_time))
print(mcmc.estimated_params['P'])
_rmse = rmse(self.p_true, mcmc.estimated_params['P'])
self.print_log('RMSE {}'.format(_rmse))
self.print_log(mcmc.summarize())
def do_experiment_pfuncs(self, ):
self.print_log('PFUNCS MODE')
self.data_model.bscc_eval_mode = BeesModel.BSCC_MODE_PFUNCS
mcmc = BayesianMcmc(self.data_model)
mcmc_chainlen = 500
mcmc.mh_params['chain_length'] = mcmc_chainlen
self.print_log(
'Start inferencing, MCMC chain len {}'.format(mcmc_chainlen))
m = self.syn_data['mult']
start_time = timeit.default_timer()
mcmc.estimate_p(m)
stop_time = timeit.default_timer()
self.print_log('MCMC estimation, ETA(secs) {}'.format(stop_time -
start_time))
print(mcmc.estimated_params['P'])
_rmse = rmse(self.p_true, mcmc.estimated_params['P'])
self.print_log('RMSE {}'.format(_rmse))
self.print_log(mcmc.summarize())
def mcmc():
return BayesianMcmc()
def create_model(self, mcmc_chainlen):
mcmc = BayesianMcmc(self.data_model)
mcmc.hyperparams['alpha'] = 2
mcmc.hyperparams['beta'] = 5
mcmc.mh_params['chain_length'] = mcmc_chainlen
return mcmc
def create_model(self, mcmc_chainlen):
mcmc = BayesianMcmc(self.data_model)
mcmc.mh_params['chain_length'] = mcmc_chainlen
return mcmc