activations='relu',
constrain_norm=constrain_norm)
return out
n = 5000
dropout_rate = min(1000. / (1000. + n), 0.5)
embedding_dropout = 0.1
embedding_l2 = 0.1
epochs = int(1500000. / float(n))
batch_size = 100
x, z, t, y, g_true = data_generator.demand(n=n,
seed=1,
ypcor=0.5,
use_images=True,
test=False)
print("Data shapes:\n\
Features:{x},\n\
Instruments:{z},\n\
Treament:{t},\n\
Response:{y}".format(**{
'x': x.shape,
'z': z.shape,
't': t.shape,
'y': y.shape
}))
# Build and fit treatment model
def datafunction(n, s, images=images, test=False):
return data_generator.demand(n=n, seed=s, ypcor=0.5, use_images=images, test=test)
'''
npr = importr('np')
y_R = robjects.FloatVector(list(y.flatten()))
(x_eval, t_eval), y_true = test_points(df, 10000)
mod = npr.npregiv(y_R,
t,
z,
x=x,
zeval=t_eval,
xeval=x_eval,
method="Tikhonov",
p=0,
optim_method="BFGS")
return ((y_true - to_array(mod.rx2('phi.eval')))**2).mean()
def prepare_file(filename):
if not os.path.exists(filename):
with open(filename, 'w') as f:
f.write('n,seed,endo,mse\n')
df = lambda n, s, test: data_generator.demand(n, s, ypcor=args.endo, test=test)
x, z, t, y, g = df(args.n_samples, args.seed, False)
mse = fit_and_evaluate(x, z, t, y, df)
DONE = True # turn off the heartbeat
prepare_file(args.results)
with open(args.results, 'a') as f:
f.write('%d,%d,%f,%f\n' % (args.n_samples, args.seed, args.endo, mse))