print("learning_rate: {}".format(learning_rate))
print("lambda_loss_amount: {}".format(lambda_loss_amount))
print("")
class EditedConfig(Config):
def __init__(self, X, Y):
super(EditedConfig, self).__init__(X, Y)
# Edit only some parameters:
self.learning_rate = learning_rate
self.lambda_loss_amount = lambda_loss_amount
# Architecture params:
self.n_layers_in_highway = n_layers_in_highway
self.n_stacked_layers = n_stacked_layers
accuracy_out, best_accuracy, f1_score_out, best_f1_score = run_with_config(
EditedConfig, X_train, y_train, X_test, y_test)
print(accuracy_out, best_accuracy, f1_score_out, best_f1_score)
with open('{}_result_opportunity_18.txt'.format(trial_name), 'a') as f:
f.write(
"""str(learning_rate)+' \t'+str(lambda_loss_amount)+' \t'+str(accuracy_out)+' \t'+str(best_accuracy)+' \t'+str(f1_score_out)+' \t'+str(best_f1_score)\n"""
)
f.write(
str(learning_rate) + ' \t' + str(lambda_loss_amount) + ' \t' +
str(accuracy_out) + ' \t' + str(best_accuracy) + ' \t' +
str(f1_score_out) + ' \t' + str(best_f1_score) + '\n\n')
print("________________________________________________________")
print("")
print("Done.")
super(EditedConfig, self).__init__(X, Y)
# Edit only some parameters:
self.learning_rate = learning_rate
self.lambda_loss_amount = lambda_loss_amount # 正则化惩罚项，值应该很小，防止过拟合
self.clip_gradients = clip_gradients
# Architecture params:
self.n_layers_in_highway = n_layers_in_highway
self.n_stacked_layers = n_stacked_layers
# # Useful catch upon looping (e.g.: not enough memory)
# try:
# accuracy_out, best_accuracy = run_with_config(EditedConfig)
# except:
# accuracy_out, best_accuracy = -1, -1
accuracy_out, best_accuracy, f1_score_out, best_f1_score = (
run_with_config(EditedConfig, X_train, y_train, X_test,
y_test))
print(accuracy_out, best_accuracy, f1_score_out, best_f1_score)
with open('{}_result_HAR_6.txt'.format(trial_name), 'a') as f:
f.write(
str(learning_rate) + ' \t' + str(lambda_loss_amount) +
' \t' + str(clip_gradients) + ' \t' + str(accuracy_out) +
' \t' + str(best_accuracy) + ' \t' + str(f1_score_out) +
' \t' + str(best_f1_score) + '\n\n')
print("________________________________________________________")
print("")
print("Done.")
super(EditedConfig, self).__init__(X, Y)
# Edit only some parameters:
self.learning_rate = learning_rate
self.lambda_loss_amount = lambda_loss_amount
self.clip_gradients = None
# Architecture params:
self.n_layers_in_highway = n_layers_in_highway
self.n_stacked_layers = n_stacked_layers
# # Useful catch upon looping (e.g.: not enough memory)
# try:
# accuracy_out, best_accuracy = run_with_config(EditedConfig)
# except:
# accuracy_out, best_accuracy = -1, -1
accuracy_out, best_accuracy, f1_score_out, best_f1_score = (
run_with_config(EditedConfig, X_train, y_train, X_test, y_test,
learning_rate))
print(accuracy_out, best_accuracy, f1_score_out, best_f1_score)
with open('{}_result_emotion_2.txt'.format(trial_name), 'a') as f:
f.write(
str(learning_rate) + ' \t' + str(lambda_loss_amount) +
' \t' + str(clip_gradients) + ' \t' + str(accuracy_out) +
' \t' + str(best_accuracy) + ' \t' + str(f1_score_out) +
' \t' + str(best_f1_score) + '\n\n')
print "________________________________________________________"
print ""
print "Done."