def fit(self, X, y):
'''
Inputs:
X: Train feature data, type: numpy array, shape: (N, num_feature)
Y: Train label data, type: numpy array, shape: (N,)
You should update the self.estimators in this function
'''
N = X.shape[0]
self.f_0 = np.average(y)
# print("f0 = ",self.f_0)
n_estimators = self.n_estimators
lr = self.learning_rate
for m in range(n_estimators):
if m == 0:
y_predict = np.repeat(self.f_0, len(y))
else:
y_predict = y_predict_new
residual = y - y_predict
regressor_m = MyDecisionTreeRegressor(min_samples_split=self.min_samples_split,max_depth=self.max_depth)
regressor_m.fit(X, residual)
y_predict_new = y_predict + lr * regressor_m.predict(X)
self.estimators[m] = regressor_m
return self.estimators
def compute_cmjs(self, residual_dict):
# For 𝑗 = 1,2, … , 𝐽𝑚, compute 𝑐𝑚𝑗
cmj_dict = copy.deepcopy(residual_dict)
self.compute_cmj_dict(cmj_dict)
cmj_tree = MyDecisionTreeRegressor(self.max_depth,
self.min_samples_split)
cmj_tree.root = cmj_dict
return cmj_tree
def fit(self, X, y):
average = np.mean(y)
self.average = average
y_predict = np.empty((X.shape[0]), dtype=np.float64)
y_predict.fill(average)
for k in range(self.n_estimators):
residual = self.negative_gradient(y, y_predict)
regression_tree = MyDecisionTreeRegressor(self.max_depth,
self.min_samples_split)
regression_tree.fit(X, residual)
y_predict += self.learning_rate * regression_tree.predict(
X).ravel()
self.estimators[k] = regression_tree
'''
def _fit_stage(self, i, X, y, y_pred):
N = X.shape[0]
residual = y - y_pred
# Induce regression tree on residuals
tree = MyDecisionTreeRegressor(
max_depth=self.max_depth, min_samples_split=self.min_samples_split)
tree.fit(X, residual)
# Compute multiplier
gamma = self.learning_rate * tree.predict(X)
# Add tree to ensemble
self.estimators[i] = tree
return y_pred + gamma
def fit(self, X, y):
'''
Inputs:
X: Train feature data, type: numpy array, shape: (N, num_feature)
Y: Train label data, type: numpy array, shape: (N,)
You should update the self.estimators in this function
'''
f = np.mean(y)
self.f0 = f
for i in range(self.n_estimators):
residual = y - f
estimator = MyDecisionTreeRegressor(
max_depth=self.max_depth,
min_samples_split=self.min_samples_split)
estimator.fit(X, residual)
f = f + self.learning_rate * np.array(estimator.predict(X))
self.estimators[i] = estimator
def fit(self, X, y):
"""
Inputs:
X: Train feature data, type: numpy array, shape: (N, num_feature)
Y: Train label data, type: numpy array, shape: (N,)
You should update the self.estimators in this function
"""
F = np.mean(y)
self.mean = F
for i in range(self.n_estimators):
# start_time = time.time ()
dtree = MyDecisionTreeRegressor(
max_depth=self.max_depth,
min_samples_split=self.min_samples_split)
dtree.fit(X, y - F)
F_pred = dtree.predict(X)
F = F + (self.learning_rate * F_pred)
self.estimators[i] = dtree
def fit(self, X, y):
'''
Inputs:
X: Train feature data, type: numpy array, shape: (N, num_feature)
Y: Train label data, type: numpy array, shape: (N,)
You should update the self.estimators in this function
'''
# np.set_printoptions(precision=17)
constant = self.get_mean(y)
self.min_y = constant
dy = y - constant
# print(dy)
for i in range(self.n_estimators):
learner_k = MyDecisionTreeRegressor(self.max_depth,
self.min_samples_split)
learner_k.fit(X, dy)
self.estimators[i] = learner_k
dy = dy - self.learning_rate * learner_k.predict(X)
return self.estimators
def fit_regression_tree(self, X, y):
# Fit a regression tree to the targets γim resulting in terminal regions 𝑅𝑚𝑗 , 𝑗 =1,2, … , 𝐽𝑚
tree = MyDecisionTreeRegressor(self.max_depth, self.min_samples_split)
tree.fit(X, y)
return tree