def test_classifier_binary():
data = load_breast_cancer()
X = data.data
y = data.target
clf_skl = KNeighborsClassifier(3) #k=3
clf_bareml = KNNClassifier(3) #k=3
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=0)
clf_skl.fit(X_train,y_train)
clf_bareml.fit(X_train,y_train)
preds_skl = clf_skl.predict(X_test).tolist()
preds_bareml = clf_bareml.predict(X_test).tolist()
# should be the same result
assert all([a == b for a, b in zip(preds_bareml, preds_skl)])
def test_gaussian():
data = load_iris()
X = data.data
y = data.target
clf_skl = GaussianNB_skl()
clf_bareml = GaussianNB()
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=0)
clf_skl.fit(X_train, y_train)
clf_bareml.fit(X_train, y_train)
preds_skl = clf_skl.predict(X_test).tolist()
preds_bareml = clf_bareml.predict(X_test).tolist()
# should be the same result
assert all([a == b for a, b in zip(preds_bareml, preds_skl)])
def test_regressor():
data = load_boston()
X = data.data
y = data.target
reg_skl = KNeighborsRegressor(3) #k=3
reg_bareml = KNNRegressor(3) #k=3
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=0)
reg_skl.fit(X_train,y_train)
reg_bareml.fit(X_train,y_train)
preds_skl = reg_skl.predict(X_test).tolist()
preds_bareml = reg_bareml.predict(X_test).tolist()
# should be the same result
assert all([a == b for a, b in zip(preds_bareml, preds_skl)])
def test_sigmoid_kernel():
data = load_boston()
X = data.data
y = data.target
reg_skl = SklKernelRidge(alpha=1.0,kernel='sigmoid')
reg_bareml = KernelRidge(alpha=1.0,kernel='sigmoid')
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=0)
reg_skl.fit(X_train,y_train)
reg_bareml.fit(X_train,y_train)
# round in 4 decimals
preds_skl = np.round(reg_skl.predict(X_test),4).tolist()
preds_bareml = np.round(reg_bareml.predict(X_test),4).tolist()
# should be the same result
assert np.allclose(preds_bareml, preds_skl)
def test_polynomial_kernel():
data = load_boston()
X = data.data
y = data.target
reg_skl = SklKernelRidge(alpha=2.0,kernel='polynomial',degree=2)
reg_bareml = KernelRidge(alpha=2.0,kernel='polynomial',degree=2)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=0)
reg_skl.fit(X_train,y_train)
reg_bareml.fit(X_train,y_train)
# with polynomial kernel, sklearn's result is not quite stable
# so check with only 1 decimal
preds_skl = np.round(reg_skl.predict(X_test),1).tolist()
preds_bareml = np.round(reg_bareml.predict(X_test),1).tolist()
# should be the same result
assert np.allclose(preds_bareml, preds_skl)
def test_gaussian_onehot():
data = load_iris()
X = data.data
y = data.target
clf_skl = GaussianNB_skl()
clf_bareml = GaussianNB()
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=0)
# one-hot encoding for bareml classifier
y_train_onehot = pd.get_dummies(y_train).values
clf_skl.fit(X_train, y_train)
clf_bareml.fit(X_train, y_train_onehot)
preds_skl = clf_skl.predict(X_test).tolist()
preds_bareml = np.argmax(clf_bareml.predict(X_test),axis=1)
# should be the same result
assert all([a == b for a, b in zip(preds_bareml, preds_skl)])
def test_ridge_gradient_descent():
data = load_boston()
X = data.data
y = data.target
reg_skl = Ridge(alpha=1.0)
reg_bareml = RidgeRegression(alpha=1.0, solver='GD')
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.3,
random_state=0)
reg_skl.fit(X_train, y_train)
reg_bareml.fit(X_train, y_train)
preds_skl = reg_skl.predict(X_test)
preds_bareml = reg_bareml.predict(X_test)
# close enough
assert all(np.abs(preds_skl - preds_bareml) < 2.0)
def test_ridge_analytical():
data = load_boston()
X = data.data
y = data.target
reg_skl = Ridge(alpha=1.0)
reg_bareml = RidgeRegression(alpha=1.0, solver='analytical')
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.3,
random_state=0)
reg_skl.fit(X_train, y_train)
reg_bareml.fit(X_train, y_train)
# round in 4 decimals
preds_skl = np.round(reg_skl.predict(X_test), 4).tolist()
preds_bareml = np.round(reg_bareml.predict(X_test), 4).tolist()
# should be the same result
assert all([a == b for a, b in zip(preds_bareml, preds_skl)])
