def transform(self, X, y):
null_dist = []
for i in range(self.n_permutation):
if self.print_progress:
progress(i, self.n_permutation)
y_perm = self.shuffle(y)
value = self.analysis.transform(X, y_perm)
null_dist.append(value)
null_dist = np.array(null_dist)
self.null_dist = null_dist
def run(self, pipeline):
results = []
for conf in self:
progress(self.i, self.n)
print str(np.float(self.i)/self.n)
pipeline.update_configuration(**conf)
res = pipeline.run()
results.append([conf, res])
self.results = results
return results
def run(self, X, y):
null_dist = []
for i in range(self.n_permutation):
if self.print_progress:
progress(i, self.n_permutation)
y_perm = self.shuffle(y)
value = self.analysis.run(X, y_perm)
null_dist.append(value)
null_dist = np.array(null_dist)
self.null_dist = null_dist
def transform(self, X, y):
"""
Runs permutations shuffiling stuff indicated in dimension variable
"""
self.null_dist = []
for i in tqdm(range(self.n_permutation), desc="permutation"):
X_, y_ = self.shuffle(X, y)
permut_value = self.analysis.transform(X_, y_)
self.null_dist.append(permut_value)
progress(i, self.n_permutation, suffix='')
self.null_dist = np.array(self.null_dist)
return self.null_dist
def run(self, X, y):
"""
The output is a vector r x n where r is the number
of repetitions of the splitting method
"""
cv = self.cross_validation
mse_ = []
set_ = []
weights_ = []
samples_ = []
i = 0
n = cv.n_iter
for train_index, test_index in cv:
i += 1
progress(i, n, suffix=' -- regression')
X_train = X[train_index]
y_train = y[train_index]
# Feature selection
self.feature_selection.run(X_train, y_train).select_first(80)
fset_ = self.feature_selection.ranking
X_train = X_train[:, fset_]
X_test = X[test_index][:,fset_]
#print X_test.shape
# We suppose only scikit-learn fit algorithms are passed!
y_predict = self.learner.fit(X_train, y_train).predict(X_test)
errors = []
for error_ in self.error_fx:
err_ = error_(y[test_index], y_predict)
errors.append(err_)
mse_.append(errors)
set_.append(fset_)
weights_.append(self.learner.coef_)
samples_.append(train_index)
self.result = [mse_, set_, weights_, samples_]
return self.result
