Varying hyperparameters
The number of iterations of training, and the size of hidden layers are two primary hyperparameters that can be varied when working with a MLP classifier. In this exercise, you will vary both separately and note how performance in terms of accuracy and AUC of the ROC curve may vary.
X_train, y_train, X_test, y_test are available in your workspace. Features have already been standardized using a StandardScaler(). pandas as pd, numpy as np are also available in your workspace.
Это упражнение является частью курса
Predicting CTR with Machine Learning in Python
Интерактивное практическое упражнение
Попробуйте выполнить это упражнение, дополнив этот пример кода.
# Loop over various max_iter configurations
max_iter_list = [10, 20, 30]
for max_iter in ____:
clf = MLPClassifier(hidden_layer_sizes = (4, ),
____ = max_iter, random_state = 0)
# Extract relevant predictions
y_score = clf.fit(____, ____).____(X_test)
y_pred = clf.fit(____, ____).____(X_test)
# Get ROC curve metrics
print("Accuracy for max_iter = %s: %s" %(
max_iter, _____(y_test, ____)))
print("AUC for max_iter = %s: %s" %(
max_iter, ____(y_test, ____[:, 1])))