def main(argv):
args = parser.parse_args(argv[1:])
# Fetch the data
(train_x, train_y), (test_x, test_y) = simu_data.load_data()
# Feature columns describe how to use the input.
my_feature_columns = []
for key in train_x.keys():
my_feature_columns.append(
tf.feature_column.numeric_column(key=key))
# Build 2 hidden layer DNN with 10, 10 units respectively.
classifier = tf.estimator.Estimator(
model_fn=gLasso_model,
params={
'feature_columns': my_feature_columns,
# Two hidden layers of 20 nodes each.
'hidden_units': [20, 20],
# The model output.
'n_response': 1,
})
# Train the Model.
classifier.train(
input_fn=lambda: simu_data.train_input_fn(
train_x, train_y, args.batch_size),
steps=args.train_steps)
# Evaluate the model.
eval_result = classifier.evaluate(
input_fn=lambda: simu_data.eval_input_fn(test_x, test_y, args.batch_size))
# extract variables from model
var_dict = dict()
for var_name in classifier.get_variable_names():
var_dict[var_name] = classifier.get_variable_value(var_name)
print('\nTest set MSE: {MSE:0.3f}\n'.format(**eval_result))
classifier = tf.estimator.Estimator(model_fn=gLasso_model,
params={
'feature_columns':
my_feature_columns,
'hidden_units': [0],
'n_response': 1,
'reg': reg,
})
classifier.train(input_fn=lambda: simu_data.train_input_fn(
train_x, train_y, 100),
steps=10000)
eval_result = classifier.evaluate(
input_fn=lambda: simu_data.eval_input_fn(test_x, test_y, 100))
var_dict = dict()
for var_name in classifier.get_variable_names():
var_dict[var_name] = classifier.get_variable_value(var_name)
v1 = np.zeros(p)
v1[0] = v1[3] = v1[6] = 1
v1 = v1.reshape((1, -1))
v2 = np.linalg.norm(var_dict["dense/kernel"], axis=1).reshape(
(1, -1))
cos_dis = cosine_similarity(v1, v2)
cos_dis = cos_dis[0][0]
spec_norm = np.linalg.norm(var_dict["dense/kernel"], 2)