def run_expt(num_graph, num_data_per_graph, n, d, s0, graph_type, sem_type,
w_ranges, noise_scale, expt_name):
os.mkdir(expt_name)
os.chmod(expt_name, 0o777)
perf = defaultdict(list)
for ii in tqdm(range(num_graph)):
B_true = utils.simulate_dag(d, s0, graph_type)
W_true = utils.simulate_parameter(B_true, w_ranges=w_ranges)
W_true_fn = os.path.join(expt_name, f'graph{ii:05}_W_true.csv')
np.savetxt(W_true_fn, W_true, delimiter=',')
for jj in range(num_data_per_graph):
X = utils.simulate_linear_sem(W_true,
n,
sem_type,
noise_scale=noise_scale)
X_fn = os.path.join(expt_name, f'graph{ii:05}_data{jj:05}_X.csv')
np.savetxt(X_fn, X, delimiter=',')
# notears
W_notears = notears.notears_linear_l1(X, lambda1=0, loss_type='l2')
assert utils.is_dag(W_notears)
W_notears_fn = os.path.join(
expt_name, f'graph{ii:05}_data{jj:05}_W_notears.csv')
np.savetxt(W_notears_fn, W_notears, delimiter=',')
# eval
B_notears = (W_notears != 0)
acc = utils.count_accuracy(B_true, B_notears)
for metric in acc:
perf[metric].append(acc[metric])
# print stats
print(expt_name)
for metric in perf:
print(metric, f'{np.mean(perf[metric]):.4f}', '+/-',
f'{np.std(perf[metric]):.4f}')
def main():
torch.set_default_dtype(torch.double)
np.set_printoptions(precision=3)
import notears.utils as ut
ut.set_random_seed(123)
n, d, s0, graph_type, sem_type = 200, 5, 9, 'ER', 'mim'
B_true = ut.simulate_dag(d, s0, graph_type)
np.savetxt('W_true.csv', B_true, delimiter=',')
X = ut.simulate_nonlinear_sem(B_true, n, sem_type)
np.savetxt('X.csv', X, delimiter=',')
model = NotearsMLP(dims=[d, 10, 1], bias=True)
W_est = notears_nonlinear(model, X, lambda1=0.01, lambda2=0.01)
assert ut.is_dag(W_est)
np.savetxt('W_est.csv', W_est, delimiter=',')
acc = ut.count_accuracy(B_true, W_est != 0)
print(acc)
def main():
torch.set_default_dtype(torch.double)
np.set_printoptions(precision=3)
import notears.utils as ut
ut.set_random_seed(123)
n, d, s0, graph_type, sem_type = 200, 5, 9, 'ER', 'mim'
ensemble_size = 7
B_true = ut.simulate_dag(d, s0, graph_type)
np.savetxt('W_true.csv', B_true, delimiter=',')
X = ut.simulate_nonlinear_sem(B_true, n, sem_type)
np.savetxt('X.csv', X, delimiter=',')
X = np.expand_dims(X, 1)
X = np.tile(X, [1, ensemble_size, 1])
model = EnsembleNotearsMLP(dims=[d, 10, 1],
ensemble_size=ensemble_size,
bias=True)
W_est = ensemble_notears_mlp(model, X, lambda1=0.01, lambda2=0.01)
if h_new > 0.25 * h:
rho *= 10
else:
break
w_est, h = w_new, h_new
alpha += rho * h
if h <= h_tol or rho >= rho_max:
break
W_est = _adj(w_est)
W_est[np.abs(W_est) < w_threshold] = 0
return W_est
if __name__ == '__main__':
from notears import utils
utils.set_random_seed(1)
n, d, s0, graph_type, sem_type = 100, 20, 20, 'ER', 'gauss'
B_true = utils.simulate_dag(d, s0, graph_type)
W_true = utils.simulate_parameter(B_true)
np.savetxt('W_true.csv', W_true, delimiter=',')
X = utils.simulate_linear_sem(W_true, n, sem_type)
np.savetxt('X.csv', X, delimiter=',')
W_est = notears_linear(X, lambda1=0.1, loss_type='l2')
assert utils.is_dag(W_est)
np.savetxt('W_est.csv', W_est, delimiter=',')
acc = utils.count_accuracy(B_true, W_est != 0)
print(acc)