def run_simulation(tmp_path, results_path):
report = SimulationReport(results_path)
samples_path = './simulation_study/samples'
for model_name in ['model1', 'model2']:
for sample_size in [5000, 10000, 20000]:
tmp = os.path.join(tmp_path, str(sample_size), str(model_name))
for sample_idx, sample in fetch_samples(model_name,
sample_size,
samples_path,
cache_dir=tmp):
bootstrap = Bootstrap(tmp, 100, '1')
smc = SMC(bootstrap,
penalty_interval=(0.01, 800),
cache_dir=None,
n_sizes=(0.3, 0.9),
alpha=0.01,
epsilon=0.01,
df_method='ct06',
perl_compatible=False,
num_cores=NUM_CORES)
smc.fit(sample)
report.add(smc, model_name, sample_size, sample_idx)
print(report.summary())
print("============== Simulation results =======================")
print(report.summary())
def test_smc_bic_pl(bootstrap, sample, tmp_path):
smc = SMC(bootstrap,
perl_compatible=True,
penalty_interval=(0.05, 100),
cache_dir=str(tmp_path))
smc.context_trees = []
smc.estimate_trees(sample)
def test_smc_caching(bootstrap, sample, tmp_path):
smc = SMC(bootstrap, penalty_interval=(0.01, 800), cache_dir=str(tmp_path))
smc2 = SMC(bootstrap,
penalty_interval=(0.01, 800),
cache_dir=str(tmp_path))
smc.fit(sample)
caching.save_cache(smc, sample)
caching.load_cache(smc2, sample)
t1 = [t.to_str() for t in smc.context_trees]
t2 = [t.to_str() for t in smc2.context_trees]
assert [t for t in t1 if t in t2] == t1
assert [t for t in t2 if t in t1] == t2
def run_smc_lcb(X):
from g4l.smc_lcb import SMC
num_cores = 1
if args.num_cores > 1:
num_cores = args.num_cores
bootstrap = Bootstrap(args.folder, args.resamples, args.renewal_point)
smc = SMC(bootstrap,
cache_dir=args.folder,
n_sizes=args.n_sizes,
alpha=args.alpha,
num_cores=num_cores)
smc.fit(X)
report(smc, X)
def run_smc_bic(X):
from g4l.smc_bic import SMC
num_cores = 1
if args.num_cores > 1:
num_cores = args.num_cores
bootstrap = Bootstrap(args.folder, args.resamples, args.renewal_point)
smc = SMC(bootstrap,
penalty_interval=tuple(args.penalty_interval),
cache_dir=args.folder,
n_sizes=args.n_sizes,
alpha=args.alpha,
epsilon=args.epsilon,
df_method=args.df,
perl_compatible=bool(args.perl_compatible),
num_cores=num_cores)
smc.fit(X)
report(smc, X)
def test_smc_bic(bootstrap, sample, tmp_path):
correct_tree = '000 100 10 1'
random.seed(12345)
smc = SMC(bootstrap, penalty_interval=(0.01, 800), cache_dir=str(tmp_path))
smc.fit(sample)
assert smc.optimal_tree.to_str(reverse=True) == correct_tree
out_path = os.path.join(str(tmp_path), 'out')
smc.save_output(sample, out_path)
html_report = open(os.path.join(out_path, 'report.html')).read()
assert html_report.find(correct_tree) >= 0