def test_add_dummy_feature(n_samples, n_features):
X = np.random.rand(n_samples, n_features)
new_skratch = skratch.add_dummy_feature(X)
new_sklearn = sklearn.add_dummy_feature(X)
assert (new_sklearn == new_skratch).all()
def fit(self, X, y):
if self.fit_intercept:
X = add_dummy_feature(X)
self.coef_ = np.linalg.inv(X.T.dot(X)).dot(X.T).dot(y)
return self.coef_
def predict(self, X, weights=None):
if self.fit_intercept:
X = add_dummy_feature(X)
if weights is None:
weights = self.coef_
return X.dot(weights)
def _loss_gradient(self, X, y):
features = X
if self.fit_intercept:
features = add_dummy_feature(features)
prediction_loss_gradient = lambda weights: (self.predict(
X, weights) - y).dot(features) / len(features)
regularization_loss_gradient = lambda weights: self.regularizer.gradient(
weights)
return lambda weights: prediction_loss_gradient(
weights) + regularization_loss_gradient(weights)
def get_X_y_weights(n_samples, n_features, fit_intercept):
X = np.random.rand(n_samples, n_features)
weights = np.random.rand(X.shape[1] + fit_intercept)
X = StandardScaler().fit_transform(X)
features = X
if fit_intercept:
features = add_dummy_feature(features)
y = features.dot(weights)
return X, y, weights
y[-1] -= 2
fig = plt.figure(figsize=(10, 10))
ax = fig.add_subplot(111)
ax.set_ylim([-2.5, 2.5])
ims = []
for weights_, new_loss in reg._fit(X, y):
X_ = np.linspace(1.5 * MIN, 1.5 * MAX, n_samples)
X_ = np.array([[x**i for i in range(1, degree + 1)] for x in X_.squeeze()])
lines = []
y_ = add_dummy_feature(X_).dot(weights_)
not_noisy = np.array(
[i for i in range(n_samples) if i not in noisy_instances])
correct, = ax.plot(X[not_noisy, 0].squeeze(), y[not_noisy], '.g')
# noise, = ax.plot(X[noisy_instances,0].squeeze(), y[noisy_instances], 'Xr')
prediction, = ax.plot(X_[:, 0].squeeze(), y_, 'b')
ax.legend([correct, prediction], ["correct", "prediction"])
lines.append(prediction)
lines.append(correct)
ims.append(lines)
anim = animation.ArtistAnimation(fig,