coursera.output('overfitting.txt', 'overfitting')
looses = {}
def plot_score(test_predictions, y_test, train_predictions, y_train, color, learning_rate):
test_loss = [log_loss(y_test, pred) for pred in test_predictions]
train_loss = [log_loss(y_train, pred) for pred in train_predictions]
plt.plot(test_loss, color, linewidth=2)
plt.plot(train_loss, color+'--', linewidth=2)
looses[learning_rate] = test_loss
plt.figure()
colors = ['r', 'g', 'b', 'c', 'm']
learn_rates = [1, 0.5, 0.3, 0.2, 0.1]
for index, learning_rate in enumerate(learn_rates):
clf.learning_rate = learning_rate
clf.fit(X_train, y_train)
test_predictions = clf.staged_predict_proba(X_test)
train_predictions = clf.staged_predict_proba(X_train)
plot_score(test_predictions, y_test, train_predictions, y_train, color=colors[index], learning_rate=learning_rate)
legends = [["Test {}".format(learn_rate), "Train {}".format(learn_rate)] for learn_rate in learn_rates]
legends = [item for sublist in legends for item in sublist]
plt.legend(legends)
plt.savefig("coursera_out/gradient_boosting.png")
min_loss_on_iteration = np.argmin(looses[0.2])
min_loss = looses[0.2][min_loss_on_iteration]
print("on iteration {} was loose {}".format(min_loss_on_iteration, min_loss))
coursera.output("min_loose_on_0.2.txt", "{:.2f} {}".format(min_loss, min_loss_on_iteration))
