def Centroid(X_train, Y_train, X_test, Y_test):
# Parameter 'shrinkage' is tuned
#Cross validation
shrinkages = np.linspace(0, 10, 100)
tuned_parameters = [{'shrink_threshold': shrinkages}]
cv = GridSearchCV(NearestCentroid(), tuned_parameters)
cv.fit(X_train, Y_train)
#Optimal parameters
print('Best Params: ')
print(cv.best_params_)
#Optimal Model
clf = NearestCentroid()
clf.set_params(shrink_threshold=cv.best_params_['shrink_threshold'])
clf.fit(X_train, Y_train)
pred = clf.predict(X_test)
test_error = mean_squared_error(Y_test, pred)
acc_score = accuracy_score(Y_test, pred)
print('Nearest Centroid Test Error: ' + str(test_error))
print('Nearest Centroid Accuracy Score: ' + str(acc_score))
print('First 10 predictions: ')
print(pred[:10])
print('First 10 actual: ')
print(Y_test[:10])
print('Centroid of each class: ')
print(clf.centroids_[0])
print('Class labels known to the classifier: ')
print(clf.classes_)
return clf, test_error, acc_score
