def test_linearsvx_loss_penalty_deprecations():
X, y = [[0.0], [1.0]], [0, 1]
msg = ("loss='%s' has been deprecated in favor of "
"loss='%s' as of 0.16. Backward compatibility"
" for the %s will be removed in %s")
# LinearSVC
# loss l1 --> hinge
assert_warns_message(DeprecationWarning,
msg % ("l1", "hinge", "loss='l1'", "1.0"),
svm.LinearSVC(loss="l1").fit, X, y)
# loss l2 --> squared_hinge
assert_warns_message(DeprecationWarning,
msg % ("l2", "squared_hinge", "loss='l2'", "1.0"),
svm.LinearSVC(loss="l2").fit, X, y)
# LinearSVR
# loss l1 --> epsilon_insensitive
assert_warns_message(
DeprecationWarning,
msg % ("l1", "epsilon_insensitive", "loss='l1'", "1.0"),
svm.LinearSVR(loss="l1").fit, X, y)
# loss l2 --> squared_epsilon_insensitive
assert_warns_message(
DeprecationWarning,
msg % ("l2", "squared_epsilon_insensitive", "loss='l2'", "1.0"),
svm.LinearSVR(loss="l2").fit, X, y)
def test_linearsvr_fit_sampleweight():
# check correct result when sample_weight is 1
# check that SVR(kernel='linear') and LinearSVC() give
# comparable results
diabetes = datasets.load_diabetes()
n_samples = len(diabetes.target)
unit_weight = np.ones(n_samples)
lsvr = svm.LinearSVR(C=1e3).fit(diabetes.data, diabetes.target,
sample_weight=unit_weight)
score1 = lsvr.score(diabetes.data, diabetes.target)
lsvr_no_weight = svm.LinearSVR(C=1e3).fit(diabetes.data, diabetes.target)
score2 = lsvr_no_weight.score(diabetes.data, diabetes.target)
assert_allclose(np.linalg.norm(lsvr.coef_),
np.linalg.norm(lsvr_no_weight.coef_), 1, 0.0001)
assert_almost_equal(score1, score2, 2)
# check that fit(X) = fit([X1, X2, X3],sample_weight = [n1, n2, n3]) where
# X = X1 repeated n1 times, X2 repeated n2 times and so forth
random_state = check_random_state(0)
random_weight = random_state.randint(0, 10, n_samples)
lsvr_unflat = svm.LinearSVR(C=1e3).fit(diabetes.data, diabetes.target,
sample_weight=random_weight)
score3 = lsvr_unflat.score(diabetes.data, diabetes.target,
sample_weight=random_weight)
X_flat = np.repeat(diabetes.data, random_weight, axis=0)
y_flat = np.repeat(diabetes.target, random_weight, axis=0)
lsvr_flat = svm.LinearSVR(C=1e3).fit(X_flat, y_flat)
score4 = lsvr_flat.score(X_flat, y_flat)
assert_almost_equal(score3, score4, 2)
def test_svr():
# Test Support Vector Regression
diabetes = datasets.load_diabetes()
for clf in (svm.NuSVR(kernel='linear', nu=.4,
C=1.0), svm.NuSVR(kernel='linear', nu=.4, C=10.),
svm.SVR(kernel='linear',
C=10.), svm.LinearSVR(C=10.), svm.LinearSVR(C=10.)):
clf.fit(diabetes.data, diabetes.target)
assert clf.score(diabetes.data, diabetes.target) > 0.02
# non-regression test; previously, BaseLibSVM would check that
# len(np.unique(y)) < 2, which must only be done for SVC
svm.SVR().fit(diabetes.data, np.ones(len(diabetes.data)))
svm.LinearSVR().fit(diabetes.data, np.ones(len(diabetes.data)))
def __init__(self):
self.clf1 = [
make_pipeline(
Imputer(),
GradientBoostingRegressor(n_estimators=5000, max_depth=8))
for _ in range(5)
]
self.clf2 = [
make_pipeline(
Imputer(strategy='median'),
ExtraTreesRegressor(n_estimators=5000,
criterion='mse',
max_depth=8,
min_samples_split=10,
min_samples_leaf=1,
min_weight_fraction_leaf=0.0,
max_features='auto',
max_leaf_nodes=None,
bootstrap=False,
oob_score=False,
n_jobs=1,
random_state=42,
verbose=0,
warm_start=True)) for _ in range(5)
]
self.clf3 = [
make_pipeline(Imputer(), svm.LinearSVR()) for _ in range(5)
]
self.clf = [linear_model.LinearRegression() for _ in range(5)]
def test_LinearSVR_C(*data):
'''
test the performance with different C
:param data: train_data,test_data, train_target, test_target
:return: None
'''
X_train, X_test, y_train, y_test = data
Cs = np.logspace(-1, 2)
train_scores = []
test_scores = []
for C in Cs:
regr = svm.LinearSVR(epsilon=0.1,
loss='squared_epsilon_insensitive',
C=C)
regr.fit(X_train, y_train)
train_scores.append(regr.score(X_train, y_train))
test_scores.append(regr.score(X_test, y_test))
fig = plt.figure()
ax = fig.add_subplot(1, 1, 1)
ax.plot(Cs, train_scores, label="Training score ", marker='+')
ax.plot(Cs, test_scores, label=" Testing score ", marker='o')
ax.set_title("LinearSVR_C ")
ax.set_xscale("log")
ax.set_xlabel(r"C")
ax.set_ylabel("score")
ax.set_ylim(-1, 1.05)
ax.legend(loc="best", framealpha=0.5)
plt.show()
def test_LinearSVR_C(*data):
"""
测试LinearSVR的预测性能随罚项系数C的变化情况
"""
train_x, test_x, train_y, test_y = data
Cs = np.logspace(-1, 2)
train_scores = []
test_scores = []
for C in Cs:
model = svm.LinearSVR(epsilon=0.1,
loss="squared_epsilon_insensitive",
C=C)
model.fit(train_x, train_y)
train_scores.append(model.score(train_x, train_y))
test_scores.append(model.score(test_x, test_y))
fig = plt.figure()
ax = fig.add_subplot(1, 1, 1)
ax.plot(Cs, train_scores, label="Training Score", marker="+")
ax.plot(Cs, test_scores, label="Testing Score", marker="o")
ax.set_title("LinearSVR_C")
ax.set_xscale("log")
ax.set_xlabel(r"C")
ax.set_ylabel("score")
ax.set_ylim(-1, 1.05)
ax.legend(loc="best")
plt.show()
def test_LinearSVR_epsilon(*data):
"""
测试LinearSVR的预测性能随eposilon参数的影响
"""
train_x, test_x, train_y, test_y = data
epsilons = np.logspace(-2, 2)
train_scores = []
test_scores = []
for epsilon in epsilons:
model = svm.LinearSVR(epsilon=epsilon,
loss="squared_epsilon_insensitive")
model.fit(train_x, train_y)
train_scores.append(model.score(train_x, train_y))
test_scores.append(model.score(test_x, test_y))
fig = plt.figure()
ax = fig.add_subplot(1, 1, 1)
ax.plot(epsilons, train_scores, label="Training Score", marker="+")
ax.plot(epsilons, test_scores, label="Testing Score", marker="o")
ax.set_title("LinearSVR_epsilon")
ax.set_xscale("log")
ax.set_xlabel(r"$\epsilon$")
ax.set_ylabel("score")
ax.set_ylim(0, 1.05)
ax.legend(loc="best")
plt.show()
def __init__(self,modeltype,PCA=None,modelparams = None):
self.information = {}
if PCA:
fitters = load(PCA)
self.pca = fitters[0]
self.scaler = fitters[1]
else:
self.pca = None
if modeltype == 'Linear': #Linear Regression
self.model = LinearRegression(n_jobs=-1)
elif modeltype == 'SVM': #Support Vector Machine
self.model= svm.SVR(cache_size=750,C=200)
elif modeltype == 'LinearSVM': #Linear SVM
self.model = svm.LinearSVR()
elif modeltype == 'SGD': #Stochastic Gradient Descent
self.model = SGDRegressor()
elif modeltype == 'MLP': #Multi-layer Perceptron
self.model = MLPRegressor(learning_rate='adaptive',max_iter=1000)
elif modeltype == 'KNN': #K Nearest Neighbour
self.model = KNeighborsRegressor(n_neighbors=2,n_jobs=-1)
elif modeltype == 'Tree': #Decision Tree
self.model = DecisionTreeRegressor()
elif modeltype == 'load': #Load a pre-existing model
pass
else: #Not supported
print('Model type not recognised')
if modelparams:
self.model.set_params(**modelparams)
def _get_base_ml_model(method):
regressor = None
if method == 'lr':
regressor = linear_model.LinearRegression()
if method == 'huber':
regressor = linear_model.HuberRegressor(max_iter=50)
regressor = multioutput.MultiOutputRegressor(regressor)
if method == 'svr':
regressor = svm.LinearSVR()
regressor = multioutput.MultiOutputRegressor(regressor)
if method == 'kr':
regressor = kernel_ridge.KernelRidge(kernel='rbf')
if method == 'rf':
regressor = ensemble.RandomForestRegressor(n_estimators=50, n_jobs=8)
if method == 'gbm':
regressor = lgb.LGBMRegressor(max_depth=20,
num_leaves=1000,
n_estimators=100,
min_child_samples=5,
random_state=42)
regressor = multioutput.MultiOutputRegressor(regressor)
if method == 'nn':
regressor = neural_network.MLPRegressor(hidden_layer_sizes=(25, 25),
early_stopping=True,
max_iter=1000000,
alpha=0.01)
return regressor
def model(train_x, train_y, test_X, flags='linear'):
y = None
if flags == 'linear':
clf = linear_model.LinearRegression()
elif flags == 'LSVR':
clf = svm.LinearSVR()
elif flags == 'SVR':
clf = svm.SVR()
elif flags == 'Ridge':
clf = linear_model.Ridge()
elif flags == 'TreeR':
clf = tree.DecisionTreeRegressor()
# Knn 不可以
elif flags == 'Knn':
clf = neighbors.KNeighborsRegressor()
elif flags == 'RandomForest':
clf = ensemble.RandomForestRegressor(n_estimators=20)
elif flags == 'Adaboost':
clf = ensemble.AdaBoostRegressor(n_estimators=50)
elif flags == 'GBRT':
clf = ensemble.GradientBoostingRegressor(n_estimators=100)
else:
pass
clf.fit(train_x, train_y)
y = clf.predict(test_X)
return y
def _estimate_model(self):
"""Estimates SVR model.
Returns
-------
model : sklearn LinearSVR or SVR model or grid search cv object
Fitted object.
"""
if self.kernel == 'linear':
self.underlying = svm.LinearSVR(**self.kwargs)
else:
if self.type == 'eps':
self.underlying = svm.SVR(kernel=self.kernel, **self.kwargs)
elif self.type == 'nu':
self.underlying = svm.NuSVR(kernel=self.kernel, **self.kwargs)
else:
raise NotImplementedError(
'Type not implemented. Choices are eps or nu.')
if self.cv_folds is not None:
model = model_selection.GridSearchCV(self.underlying,
self.parameters,
cv=self.cv_folds,
scoring=self.score)
else:
model = self.underlying
model.fit(self.x_train, self.y_train)
return model
def linear_svr_c(*data):
x_train, x_test, y_train, y_test = data
cs = np.logspace(-1, 3)
train_scores = []
test_scores = []
for c in cs:
svr = svm.LinearSVR(epsilon=0.1,
loss="squared_epsilon_insensitive",
C=c)
svr.fit(x_train, y_train)
train_scores.append(svr.score(x_train, y_train))
test_scores.append(svr.score(x_test, y_test))
fig = plt.figure()
ax = fig.add_subplot(1, 1, 1)
ax.plot(cs, train_scores, label="training score", marker="+")
ax.plot(cs, test_scores, label="testing score", marker="o")
ax.set_title("svr c")
ax.set_xscale("log")
ax.set_xlabel(r"C")
ax.set_ylabel("score")
ax.set_ylim(-1, 1.05)
ax.legend(loc="best", framealpha=0.5)
plt.show()
def regression():
new_url = 'https://goo.gl/sXleFv'
new_columns = np.array([
'CRIM', 'ZN', 'INDUS', 'CHAS', 'NOX', 'RM', 'AGE', 'DIS', 'RAD', 'TAX',
'PTRATIO', 'B', 'LSTAT', 'MEDV'
])
dataframe = pd.read_csv(new_url, delim_whitespace=True, names=new_columns)
array = dataframe.values
X = array[:, 0:13]
Y = array[:, 13]
k_fold = model_selection.KFold(n_splits=10, random_state=7)
models = np.empty([6, 2], dtype='object')
models[0] = ['K Nearest Neighbors', neighbors.KNeighborsRegressor()]
models[1] = ['Linear Regression', linear_model.LinearRegression()]
models[2] = ['Ridge Regression', linear_model.Ridge()]
models[3] = ['Support Vector Regressor', svm.LinearSVR()]
models[4] = ['Random Forest Regressor', ensemble.RandomForestRegressor()]
models[5] = [
'Gradient Boosted Trees',
ensemble.GradientBoostingRegressor()
]
for name, model in models:
# Different model metrics
for scoring in ('neg_mean_squared_error', 'explained_variance'):
cross_validation(name, model, X, Y, scoring)
def test_LinearSVR_C(*data):
'''
测试 LinearSVR 的预测性能随 C 参数的影响
:param data: 可变参数。它是一个元组，这里要求其元素依次为：训练样本集、测试样本集、训练样本的值、测试样本的值
:return: None
'''
X_train, X_test, y_train, y_test = data
Cs = np.logspace(-1, 2)
train_scores = []
test_scores = []
for C in Cs:
regr = svm.LinearSVR(epsilon=0.1,
loss='squared_epsilon_insensitive',
C=C)
regr.fit(X_train, y_train)
train_scores.append(regr.score(X_train, y_train))
test_scores.append(regr.score(X_test, y_test))
fig = plt.figure()
ax = fig.add_subplot(1, 1, 1)
ax.plot(Cs, train_scores, label="Training score ", marker='+')
ax.plot(Cs, test_scores, label="Testing score ", marker='o')
ax.set_title("LinearSVR_C ")
ax.set_xscale("log")
ax.set_xlabel(r"C")
ax.set_ylabel("score")
ax.set_ylim(-1, 1.05)
ax.legend(loc="best", framealpha=0.5)
plt.show()
def predict(data, priceToPredict):
openingPriceTrain, openingPriceTest, closingPriceTrain, closingPriceTest = \
data["openingPriceTrain"], data["openingPriceTest"], data["closingPriceTrain"], data["closingPriceTest"]
clf = svm.LinearSVR()
clf.fit(openingPriceTrain, closingPriceTrain)
predicted2 = clf.predict(openingPriceTest)
score = clf.fit(openingPriceTrain,
closingPriceTrain).score(openingPriceTest,
closingPriceTest)
# print(score)
fig, ax = plotter.subplots()
ax.scatter(openingPriceTrain, closingPriceTrain)
ax.set_ylabel('Predicted SVM')
ax.scatter(closingPriceTest, clf.predict(openingPriceTest))
ax.set_xlabel('Measured')
ax.set_ylabel('Predicted')
# plotter.show()
closingPriceTestArray = np.reshape(closingPriceTest, -1)
clfpr = clf.predict(openingPriceTest)
predictedArray = np.reshape(clfpr, -1)
print(pearsonr(closingPriceTestArray, predictedArray))
openingPriceToPredict = np.array([priceToPredict])
print(clf.predict(openingPriceToPredict))
return clf.predict(np.array([openingPriceToPredict]))
def multiple_comparision():
ln_predictor = linear_model.LinearRegression()
svm_predictor = svm.LinearSVR()
tree_predictor = tree.DecisionTreeRegressor()
kernel_predictor = KernelRidge(alpha=1.0, gamma=1.0, kernel='rbf')
C = 15
printflag = False
predictor_dict = {
'Linear': ln_predictor,
'SVR': svm_predictor,
'DT': tree_predictor,
'RBF Kernel': kernel_predictor
}
gamma_list = [0.01]
max_iter_list = [5]
results_dict = {}
# For each model, train a Pareto curve
for max_iter in max_iter_list:
for curr_predictor in predictor_dict.keys():
print('Curr Predictor: ')
print(curr_predictor)
predictor = predictor_dict[curr_predictor]
fair_clf = Model(C=C,
printflag=printflag,
gamma=1,
predictor=predictor,
max_iters=max_iter)
print(fair_clf.predictor)
all_errors, all_fp = fair_clf.pareto(X, X_prime, y, gamma_list)
results_dict[curr_predictor] = {
'Errors': all_errors,
'FP_disp': all_fp
}
print(results_dict)
def run_sklearn():
skm = sk.LinearSVR(loss=loss,
epsilon=eps,
max_iter=skit,
dual=skdual)
skm.fit(X_train, y_train)
return skm.score(X_test, y_test)
def svm(self, type_, epsilon=0.0, penalty=1.0, tol=0.0001):
"""
---------------------------------------------
Regression using Support Vector Machines (SVM)
---------------------------------------------
Parameters:
epsilon: Parameter in loss function. Defines margin where no penalty is given to errors.
penalty: L2-penalty for error term. The larger, the less regularisation is used.
tol: Tolerance for stopping criteria
loss: Set to epsilon_insensitive, standard SVR
"""
self.penalty = penalty
self.eps = epsilon
self.tol = tol
self.clf = svm.LinearSVR(epsilon=self.eps,
tol=self.tol,
C=self.penalty,
loss='epsilon_insensitive',
fit_intercept=False,
max_iter=10e5)
fit = self.clf.fit(self.X, self.y)
self.weights = self.clf.coef_
pred = Regression.predict(self, self.X)
MSE = mean_squared_error(self.y, pred)
return MSE, self.clf.score(self.X, self.y)
def test_linear_svx_uppercase_loss_penalty():
# Check if Upper case notation is supported by _fit_liblinear
# which is called by fit
X, y = [[0.0], [1.0]], [0, 1]
msg = ("loss='%s' has been deprecated in favor of "
"loss='%s' as of 0.16. Backward compatibility"
" for the uppercase notation will be removed in %s")
# loss SQUARED_hinge --> squared_hinge
assert_warns_message(DeprecationWarning,
msg % ("SQUARED_hinge", "squared_hinge", "0.18"),
svm.LinearSVC(loss="SQUARED_hinge").fit, X, y)
# penalty L2 --> l2
assert_warns_message(DeprecationWarning,
msg.replace("loss", "penalty")
% ("L2", "l2", "0.18"),
svm.LinearSVC(penalty="L2").fit, X, y)
# loss EPSILON_INSENSITIVE --> epsilon_insensitive
assert_warns_message(DeprecationWarning,
msg % ("EPSILON_INSENSITIVE", "epsilon_insensitive",
"0.18"),
svm.LinearSVR(loss="EPSILON_INSENSITIVE").fit, X, y)
def __init__(self, trainer=svr.LinearSVR(), error_fx=sklm.r2_score):
#super(RegressionMeasure, self).__init__()
Measure.__init__(self)
self.trainer = trainer
self.fx = error_fx
self.mse = sklm.mean_squared_error
def multiple_pareto():
gamma_list = [0.002, 0.005, 0.01, 0.02, 0.05, 0.1]
train_size = X.shape[0]
X_train = X.iloc[:train_size]
X_prime_train = X_prime.iloc[:train_size]
y_train = y.iloc[:train_size]
ln_predictor = linear_model.LinearRegression()
svm_predictor = svm.LinearSVR()
tree_predictor = tree.DecisionTreeRegressor(max_depth=3)
kernel_predictor = KernelRidge(alpha=1.0, gamma=1.0, kernel='rbf')
predictor_dict = {
'Linear': {
'predictor': ln_predictor,
'iters': 100
},
'SVR': {
'predictor': svm_predictor,
'iters': 10
},
'DT': {
'predictor': tree_predictor,
'iters': 100
}
}
results_dict = {}
for pred in predictor_dict:
print('Curr Predictor: {}'.format(pred))
predictor = predictor_dict[pred]['predictor']
max_iters = predictor_dict[pred]['iters']
fair_clf = Model(C=100,
printflag=True,
gamma=1,
predictor=predictor,
max_iters=max_iters)
fair_clf.set_options(max_iters=max_iters)
errors, fp_violations, fn_violations = fair_clf.pareto(
X_train, X_prime_train, y_train, gamma_list)
results_dict[pred] = {
'errors': errors,
'fp_violations': fp_violations,
'fn_violations': fn_violations
}
plt.plot(errors, fp_violations, label=pred)
pickle.dump(
results_dict,
open(
'results_dict_' + str(gamma_list) + '_gammas' + str(gamma_list) +
'.pkl', 'wb'))
plt.xlabel('Error')
plt.ylabel('Unfairness')
plt.legend()
plt.title('Error vs. Unfairness\n(Communities & Crime Dataset)')
plt.show()
def train(self, results, features, hyperparams, feature_names):
if self.is_classifier:
model = svm.LinearSVC(C=hyperparams.get(self.C_VAL))
else:
model = svm.LinearSVR(C=hyperparams.get(self.C_VAL), epsilon=hyperparams.get(self.EPSILON))
model.fit(features, results)
self.log.debug("Successful creation of Linear Support Vector Machine model: %s\n", model)
return model
def test_sk_LinearSVR():
print("Testing sklearn, LinearSVR...")
mod = svm.LinearSVR()
X, y = iris_data
mod.fit(X, y)
docs = {'name': "LinearSVR test"}
fv = X[0, :]
upload(mod, fv, docs)
def Train(complexity, trainData, trainLabels):
seed = 1
reg = svm.LinearSVR(C=complexity,
epsilon=0.1,
random_state=seed,
loss='squared_epsilon_insensitive')
reg.fit(trainData, trainLabels)
return reg
def linear_svr(*data):
x_train, x_test, y_train, y_test = data
svr = svm.LinearSVR()
svr.fit(x_train, y_train)
print(svr.coef_)
print(svr.intercept_)
print(svr.score(x_train, y_train))
print(svr.score(x_test, y_test))
def reset(self):
if self.kernel == 'linear':
self.clf = svm.LinearSVR(random_state=self.seed)
else:
self.clf = svm.SVR(gamma='auto',
kernel=self.kernel,
degree=self.poly_degree,
random_state=self.seed)
def calculate_compartment_fraction(structure1,
structure2,
path1,
path2,
size1=None,
size2=None):
#compartments
contacts1 = matFromBed(path1, size1, structure1)
contacts2 = matFromBed(path2, size2, structure2)
compartments1 = np.array(get_compartments(contacts1, structure1))
compartments2 = np.array(get_compartments(contacts2, structure2))
r, p = st.pearsonr(compartments1, compartments2)
if r < 0:
compartments2 = -compartments2
#SVR
coords1 = structure1.getCoords()
coords2 = structure2.getCoords()
coords = np.concatenate((coords1, coords2))
compartments = np.concatenate((compartments1, compartments2))
clf = svm.LinearSVR()
clf.fit(coords, compartments)
coef = clf.coef_
transformed_coords1 = np.array(change_coordinate_system(coef, coords1))
transformed_coords2 = np.array(change_coordinate_system(coef, coords2))
x_diffs = transformed_coords1[:, 0] - transformed_coords2[:, 0]
y_diffs = transformed_coords1[:, 1] - transformed_coords2[:, 1]
z_diffs = transformed_coords1[:, 2] - transformed_coords2[:, 2]
#axis lengths
centroid1 = np.mean(transformed_coords1, axis=0)
centroid2 = np.mean(transformed_coords2, axis=0)
x_length1 = np.mean(
[np.abs(coord1[0] - centroid1[0]) for coord1 in transformed_coords1])
y_length1 = np.mean(
[np.abs(coord1[1] - centroid1[1]) for coord1 in transformed_coords1])
z_length1 = np.mean(
[np.abs(coord1[2] - centroid1[2]) for coord1 in transformed_coords1])
x_length2 = np.mean(
[np.abs(coord2[0] - centroid2[0]) for coord2 in transformed_coords2])
y_length2 = np.mean(
[np.abs(coord2[1] - centroid2[1]) for coord2 in transformed_coords2])
z_length2 = np.mean(
[np.abs(coord2[2] - centroid2[2]) for coord2 in transformed_coords2])
x_length = np.mean((x_length1, x_length2))
y_length = np.mean((y_length1, y_length2))
z_length = np.mean((z_length1, z_length2))
x_mean = np.mean(np.abs(x_diffs)) / x_length
y_mean = np.mean(np.abs(y_diffs)) / y_length
z_mean = np.mean(np.abs(z_diffs)) / z_length
return z_mean / (x_mean + y_mean + z_mean)
def LinearSVR(self,
optimizer=None,
param_grid=None,
scoring=None,
fit_params=None,
cv=None):
"""Creates a linear support vector machine regression model estimator.
:param optimizer: the parameter search and optimization method. default is a sklearn.model_selection.GridSearchCV instance
:param param_grid: dictionary with hyperparameter names as keys and lists of hyperparameter settings to try in grid search
:param scoring: the model performance metric to optimize. accepts string values and sklear.metrics.* instances
:param fit_params: parameters to pass to the `fit` method of the estimator
:param cv: determines the cross-validation splitting strategy. accepts inputs to sklearn.model_selection.GridSearchCV
:return None: updates the `tuner` object with the passed parameters and runs a grid search using the LinearSVR estimator
"""
# check if the optimizer has changed, otherwise use default
if optimizer is not None:
self.optimizer = optimizer
# check if the parameter grid has been set, otherwise set defaults
if param_grid is None:
if self.param_grid is None:
param_grid = {
"C": (1e-2, 1e-1, 1e0, 1e1),
"loss":
("epsilon_insensitive", "squared_epsilon_insensitive"),
"epsilon": (0, 0.01, 0.1),
"dual": (False),
"tol": (1e-3, 1e-4, 1e-5),
"fit_intercept": (True, False),
}
self.param_grid = param_grid
else:
self.param_grid = param_grid
# set the scoring function
if scoring is None:
if self.scoring is None:
scoring = _metrics.explained_variance_score
self.scoring = scoring
else:
self.scoring = scoring
# set the default fit parameters
if fit_params is not None:
self.fit_params = fit_params
# set the cross validation strategy
if cv is None:
if self.cv is None:
cv = _model_selection.StratifiedKFold(n_splits=self.n_splits)
self.cv = cv
else:
self.cv = cv
# create the estimator and run the grid search
estimator = _svm.LinearSVR()
self.run_gs(estimator)
def test_LinearSVR_loss(*data):
x_train, x_test, y_train, y_test = data
losses = ['epsilon_insensitive', 'squared_epsilon_insensitive']
for loss in losses:
regr = svm.LinearSVR(loss=loss)
regr.fit(x_train, y_train)
print("loss: %s" % loss)
print("Coefficients:%s, intercept:%s" % (regr.coef_, regr.intercept_))
print("Score: %.2f" % regr.score(x_test, y_test))
def build_model(args, C, seed):
if args.dc_tree:
model = DecisionTreeRegressor(random_state=seed)
elif args.nn_radius:
model = RadiusNeighborsRegressor(radius=1.0)
else:
model = svm.LinearSVR(C=complexities[comp], random_state=seed)
return model