def run():
# Load image paths in the dataset
training_image_path_list, testing_image_path_list = load_image_path_list()
# Load features
training_image_feature_dict = load_features(training_image_path_list)
testing_image_feature_dict = load_features(testing_image_path_list)
# Load training labels
training_names, training_labels, testing_names = load_csv_files()
# Convert data to suitable form for training/testing phase
X_train = get_attributes(training_image_feature_dict, training_names)
Y_train = training_labels
X_test = get_attributes(testing_image_feature_dict, testing_names)
# Generate prediction list
prediction_list = []
for trial_index in range(11):
print("Working on trial NO.{:d}".format(trial_index + 1))
current_prediction = keras_NN.generate_prediction(X_train, Y_train, X_test)
prediction_list.append(current_prediction)
# Generate ensemble prediction
ensemble_prediction, _ = stats.mode(prediction_list)
ensemble_prediction = np.squeeze(ensemble_prediction)
# Create submission file
submission_file_name = "Aurora_" + str(int(time.time())) + ".csv"
file_content = pd.DataFrame({"Id": testing_names, "Prediction": ensemble_prediction})
file_content.to_csv(submission_file_name, index=False, header=True)
print("All done!")
def run():
# Load image paths in the dataset
training_image_path_list, testing_image_path_list = load_image_path_list()
# Load features
training_image_feature_dict = load_features(training_image_path_list)
testing_image_feature_dict = load_features(testing_image_path_list)
# Load training labels
training_names, training_labels, testing_names = load_csv_files()
# Convert data to suitable form for training/testing phase
X_train = get_attributes(training_image_feature_dict, training_names)
Y_train = training_labels
X_test = get_attributes(testing_image_feature_dict, testing_names)
# Generate prediction list
prediction_list = []
for trial_index in range(11):
print("Working on trial NO.{:d}".format(trial_index + 1))
current_prediction = keras_NN.generate_prediction(
X_train, Y_train, X_test)
prediction_list.append(current_prediction)
# Generate ensemble prediction
ensemble_prediction, _ = stats.mode(prediction_list)
ensemble_prediction = np.squeeze(ensemble_prediction)
# Create submission file
submission_file_name = "Aurora_" + str(int(time.time())) + ".csv"
file_content = pd.DataFrame({
"Id": testing_names,
"Prediction": ensemble_prediction
})
file_content.to_csv(submission_file_name, index=False, header=True)
print("All done!")
import time
# Read data set from file
X_train = pd.read_csv("./input/X_train.csv", skiprows=0).as_matrix()[:, 1:]
multi_Y_train = pd.read_csv("./input/y_train.csv", skiprows=0).as_matrix()[:,
1:]
X_test = pd.read_csv("./input/X_test.csv", skiprows=0).as_matrix()[:, 1:]
multi_Y_test = []
# Generate prediction for each angle
for current_column in range(multi_Y_train.shape[1]):
Y_train = multi_Y_train[:, current_column]
prediction = keras_NN.generate_prediction(X_train,
Y_train,
X_test,
True,
layer_size=512,
layer_num=5,
nb_epoch=100)
multi_Y_test.append(np.reshape(prediction, (-1, 1)))
# Create submission file
multi_Y_test = np.hstack(multi_Y_test)
ID = np.arange(X_train.shape[0] + 1,
X_train.shape[0] + multi_Y_test.shape[0] + 1)
submission_file_name = "Aurora_" + str(int(time.time())) + ".csv"
submission_file_DataFrame = pd.DataFrame({
"Id": ID,
"Angle1": multi_Y_test[:, 0],
"Angle2": multi_Y_test[:, 1]
})
import keras_NN
import numpy as np
import pandas as pd
import time
# Read data set from file
X_train = pd.read_csv("./input/X_train.csv", skiprows=0).as_matrix()[:, 1:]
multi_Y_train = pd.read_csv("./input/y_train.csv", skiprows=0).as_matrix()[:, 1:]
X_test = pd.read_csv("./input/X_test.csv", skiprows=0).as_matrix()[:, 1:]
multi_Y_test = []
# Generate prediction for each angle
for current_column in range(multi_Y_train.shape[1]):
Y_train = multi_Y_train[:, current_column]
prediction = keras_NN.generate_prediction(X_train, Y_train, X_test, True, layer_size=512, layer_num=5, nb_epoch=100)
multi_Y_test.append(np.reshape(prediction, (-1, 1)))
# Create submission file
multi_Y_test = np.hstack(multi_Y_test)
ID = np.arange(X_train.shape[0] + 1, X_train.shape[0] + multi_Y_test.shape[0] + 1)
submission_file_name = "Aurora_" + str(int(time.time())) + ".csv"
submission_file_DataFrame = pd.DataFrame({"Id": ID, "Angle1": multi_Y_test[:, 0], "Angle2": multi_Y_test[:, 1]})
submission_file_DataFrame.to_csv(submission_file_name, index=False, header=True)
print("All done!")