def test_fit(self):
clf = Perceptron(weight_initialization='zeros', shuffle=False)
X, y = get_data('table2')
clf.fit(X, y)
y_hat = clf.predict(X)
score = accuracy_score(y, y_hat)
assert_almost_equal(score, 0.8333333)
def test_fit(self):
clf = LinearGradientDescent()
X, y = get_data('table2')
clf.fit(X, y)
y_hat = clf.predict(X)
score = accuracy_score(y, y_hat)
assert_almost_equal(score, 0.833333333)
def test_fit_madelon(self):
clf = SVM(shuffle=False, epochs=1, weight_initialization='zeros')
X, y = get_data('madelon')
clf.fit(X, y)
y_hat = clf.predict(X)
score = accuracy_score(y, y_hat)
assert_almost_equal(score, 0.5505, decimal=4)
def test_fit_madelon(self):
np.random.seed(0)
clf = SVMRandomForest(N=5, m=20)
X, y = get_data('madelon')
clf.fit(X, y)
y_hat = clf.predict(X)
score = accuracy_score(y, y_hat)
assert_almost_equal(score, -1)
def test_fit_handwriting(self):
np.random.seed(0)
clf = SVMRandomForest(N=5, m=100)
X, y = get_data('handwriting')
clf.fit(X, y)
y_hat = clf.predict(X)
score = accuracy_score(y, y_hat)
assert_almost_equal(score, 1)
def test_numerical_data_works_with_missing_method_choices(self):
X, y = get_data('blobs')
clf = DecisionTree(missing_method='majority_value')
clf.fit(X, y)
y_hat = clf.predict(X)
score = accuracy_score(y, y_hat)
self.assertTrue(score > 0.97)
def test_numerical_data(self):
X, y = get_data('blobs')
clf = DecisionTree()
clf.fit(X, y)
y_hat = clf.predict(X)
score = accuracy_score(y, y_hat)
self.assertTrue(score > 0.97)
def test_on_numeric_madelon(self):
X, y = get_data('madelon')
X_test, y_test = get_data('madelon', test=True)
clf = DecisionTree()
clf.fit(X, y)
y_hat = clf.predict(X_test)
score = accuracy_score(y_test, y_hat)
self.assertTrue(score > 0.97)
def test_setting_a(self):
X, y = get_data('mushroom', setting='SettingA')
X_test, y_test = get_data('mushroom', setting='SettingA', test=True)
clf = DecisionTree()
clf.fit(X, y)
y_hat = clf.predict(X_test)
score = accuracy_score(y_test, y_hat)
self.assertEqual(score, 1)
# how does it compare to sklearn
clf = DecisionTreeClassifier()
Xt = pd.get_dummies(pd.DataFrame(X)).values
Xt_test = pd.get_dummies(pd.DataFrame(X_test)).values
clf.fit(Xt, y)
y_hat = clf.predict(Xt_test)
score = accuracy_score(y_test, y_hat)
self.assertEqual(score, 1)
def test_stuff(self):
return # TODO
# TODO how do you use the same base classifier over & over?
# nothing makes it any different, so how do you get it
# to select different things...
X,y = get_data('table2')
clf = AdaBoost([
DecisionTree(max_depth=1),
DecisionTree(max_depth=2),
DecisionTree(max_depth=3),
DecisionTree(max_depth=1, best_attr_method='gini'),
DecisionTree(max_depth=2, best_attr_method='gini'),
DecisionTree(max_depth=3, best_attr_method='gini'),
], T=5)
clf.fit(X, y)
y_hat = clf.predict(X)
score = accuracy_score(y, y_hat)
self.assertEqual(score, 1)
def test_hw_example(self):
classifiers = [
RuleClassifier(lambda x: np.sign(x[:,0])),
RuleClassifier(lambda x: np.sign(x[:,0] - 2)),
RuleClassifier(lambda x: -np.sign(x[:,0])),
RuleClassifier(lambda x: -np.sign(x[:,1])),
]
clf = AdaBoost(classifiers, T=4)
X = np.array([
[1,1],
[1,-1],
[-1,-1],
[-1,1],
])
y = np.array([-1,1,-1,-1])
clf.fit(X, y)
y_hat = clf.predict(X)
score = accuracy_score(y, y_hat)
self.assertEqual(score, 1)
def test_classifier(self):
classifiers=[
RuleClassifier(lambda x: np.sign(1.5 - x[:, 0])),
RuleClassifier(lambda x: np.sign(4.5 - x[:, 0])),
RuleClassifier(lambda x: np.sign(x[:, 1] - 5)),
]
clf = AdaBoost(classifiers, T=5)
X = np.array([
[1, 2],
[1, 4],
[2.5, 5.5],
[3.5, 6.5],
[4, 5.4],
[2, 1],
[2, 4],
[3.5, 3.5],
[5, 2],
[5, 5.5],
])
y = np.array([1,1,1,1,1,-1,-1,-1,-1,-1])
clf.fit(X, y)
y_hat = clf.predict(X)
score = accuracy_score(y, y_hat)
self.assertEqual(score, 1)
def test_accuracy(self):
score = accuracy_score(
np.array([0, 0, 0, 0, 0, 1, 1, 1, 1, 1]),
np.array([1, 0, 1, 0, 1, 0, 1, 0, 1, 0]),
)
assert_almost_equal(score, 0.4)