def bio_classification_report(y_true, y_pred):
"""
Classification report for a list of BIO-encoded sequences.
It computes token-level metrics and discards "O" labels.
Note that it requires scikit-learn 0.15+ (or a version from github master)
to calculate averages properly!
Note: This function was copied from
http://nbviewer.ipython.org/github/tpeng/python-crfsuite/blob/master/examples/CoNLL%202002.ipynb
Args:
y_true: True labels, list of strings
y_pred: Predicted labels, list of strings
Returns:
classification report as string
"""
lbin = LabelBinarizer()
y_true_combined = lbin.fit_transform(list(chain.from_iterable(y_true)))
y_pred_combined = lbin.transform(list(chain.from_iterable(y_pred)))
#tagset = set(lbin.classes_) - {NO_NE_LABEL}
tagset = set(lbin.classes_)
tagset = sorted(tagset, key=lambda tag: tag.split('-', 1)[::-1])
class_indices = {cls: idx for idx, cls in enumerate(lbin.classes_)}
return classification_report(
y_true_combined,
y_pred_combined,
labels=[class_indices[cls] for cls in tagset],
target_names=tagset,
)
class MLPClassifier(BaseMLP, ClassifierMixin):
""" Multilayer Perceptron Classifier.
Uses a neural network with one hidden layer.
Parameters
----------
Attributes
----------
Notes
-----
References
----------"""
def __init__(
self, n_hidden=200, lr=0.1, l2decay=0, loss="cross_entropy", output_layer="softmax", batch_size=100, verbose=0
):
super(MLPClassifier, self).__init__(n_hidden, lr, l2decay, loss, output_layer, batch_size, verbose)
def fit(self, X, y, max_epochs=10, shuffle_data=False):
self.lb = LabelBinarizer()
one_hot_labels = self.lb.fit_transform(y)
super(MLPClassifier, self).fit(X, one_hot_labels, max_epochs, shuffle_data)
return self
def predict(self, X):
prediction = super(MLPClassifier, self).predict(X)
return self.lb.inverse_transform(prediction)
def test_multinomial_loss():
# test if the multinomial loss and gradient computations are consistent
X, y = iris.data, iris.target.astype(np.float64)
n_samples, n_features = X.shape
n_classes = len(np.unique(y))
rng = check_random_state(42)
weights = rng.randn(n_features, n_classes)
intercept = rng.randn(n_classes)
sample_weights = rng.randn(n_samples)
np.abs(sample_weights, sample_weights)
# compute loss and gradient like in multinomial SAG
dataset, _ = make_dataset(X, y, sample_weights, random_state=42)
loss_1, grad_1 = _multinomial_grad_loss_all_samples(dataset, weights,
intercept, n_samples,
n_features, n_classes)
# compute loss and gradient like in multinomial LogisticRegression
lbin = LabelBinarizer()
Y_bin = lbin.fit_transform(y)
weights_intercept = np.vstack((weights, intercept)).T.ravel()
loss_2, grad_2, _ = _multinomial_loss_grad(weights_intercept, X, Y_bin,
0.0, sample_weights)
grad_2 = grad_2.reshape(n_classes, -1)
grad_2 = grad_2[:, :-1].T
# comparison
assert_array_almost_equal(grad_1, grad_2)
assert_almost_equal(loss_1, loss_2)
class BinaryRelevanceClassifier(BaseEstimator, ClassifierMixin):
def __init__(self, estimator):
self.estimator = estimator
def fit(self, X, Y):
# binarize labels
self.bl = LabelBinarizer()
Y = self.bl.fit_transform(Y)
self.classes_ = self.bl.classes_
# create an estimator for each label
self.estimators_ = []
for i in xrange(self.bl.classes_.shape[0]):
estimator = clone(self.estimator)
estimator.fit(X, Y[:, i])
self.estimators_.append(estimator)
def predict(self, X):
self._check_is_fitted()
X = np.atleast_2d(X)
Y = np.empty((X.shape[0], self.classes_.shape[0]))
for i, estimator in enumerate(self.estimators_):
Y[:, i] = estimator.predict(X).T
return self.bl.inverse_transform(Y)
def _check_is_fitted(self):
if not hasattr(self, "estimators_"):
raise ValueError("The object hasn't been fitted yet!")
def test_multinomial_loss_ground_truth():
# n_samples, n_features, n_classes = 4, 2, 3
n_classes = 3
X = np.array([[1.1, 2.2], [2.2, -4.4], [3.3, -2.2], [1.1, 1.1]])
y = np.array([0, 1, 2, 0])
lbin = LabelBinarizer()
Y_bin = lbin.fit_transform(y)
weights = np.array([[0.1, 0.2, 0.3], [1.1, 1.2, -1.3]])
intercept = np.array([1., 0, -.2])
sample_weights = np.array([0.8, 1, 1, 0.8])
prediction = np.dot(X, weights) + intercept
logsumexp_prediction = logsumexp(prediction, axis=1)
p = prediction - logsumexp_prediction[:, np.newaxis]
loss_1 = -(sample_weights[:, np.newaxis] * p * Y_bin).sum()
diff = sample_weights[:, np.newaxis] * (np.exp(p) - Y_bin)
grad_1 = np.dot(X.T, diff)
weights_intercept = np.vstack((weights, intercept)).T.ravel()
loss_2, grad_2, _ = _multinomial_loss_grad(weights_intercept, X, Y_bin,
0.0, sample_weights)
grad_2 = grad_2.reshape(n_classes, -1)
grad_2 = grad_2[:, :-1].T
assert_almost_equal(loss_1, loss_2)
assert_array_almost_equal(grad_1, grad_2)
# ground truth
loss_gt = 11.680360354325961
grad_gt = np.array([[-0.557487, -1.619151, +2.176638],
[-0.903942, +5.258745, -4.354803]])
assert_almost_equal(loss_1, loss_gt)
assert_array_almost_equal(grad_1, grad_gt)
def chi2_contingency_matrix(X_train, y_train):
X = X_train.copy()
X.data = np.ones_like(X.data)
X = check_array(X, accept_sparse='csr')
if np.any((X.data if issparse(X) else X) < 0):
raise ValueError("Input X must be non-negative.")
Y = LabelBinarizer().fit_transform(y_train)
if Y.shape[1] == 1:
Y = np.append(1 - Y, Y, axis=1)
observed = safe_sparse_dot(Y.T, X) # n_classes * n_features
# feature_count = check_array(X.sum(axis=0))
# class_prob = check_array(Y.mean(axis=0))
feature_count = X.sum(axis=0).reshape(1, -1)
class_prob = Y.mean(axis=0).reshape(1, -1)
expected = np.dot(class_prob.T, feature_count)
observed = np.asarray(observed, dtype=np.float64)
k = len(observed)
# Reuse observed for chi-squared statistics
contingency_matrix = observed
contingency_matrix -= expected
contingency_matrix **= 2
expected[expected == 0.0] = 1.0
contingency_matrix /= expected
# weights = contingency_matrix.max(axis=0)
return contingency_matrix
class GBClassifier(_BaseGB, ClassifierMixin):
def __init__(self, estimator, n_estimators=100,
step_size="line_search", learning_rate=0.1,
loss="squared_hinge", subsample=1.0,
callback=None, random_state=None):
self.estimator = estimator
self.n_estimators = n_estimators
self.step_size = step_size
self.learning_rate = learning_rate
self.loss = loss
self.subsample = subsample
self.callback = callback
self.random_state = random_state
def _get_loss(self):
losses = dict(squared_hinge=_SquaredHingeLoss(),
log=_LogLoss())
return losses[self.loss]
def fit(self, X, y):
self._lb = LabelBinarizer(neg_label=-1)
Y = self._lb.fit_transform(y)
return super(GBClassifier, self).fit(X, Y)
def predict(self, X):
pred = self.decision_function(X)
return self._lb.inverse_transform(pred)
def bio_classification_report(y_true, y_pred):
"""
Classification report for a list of BIO-encoded sequences.
It computes token-level metrics and discards "O" labels.
Note that it requires scikit-learn 0.15+ (or a version from
github master) to calculate averages properly!
"""
lb = LabelBinarizer()
y_true_combined = lb.fit_transform(list(chain.from_iterable(y_true)))
y_pred_combined = lb.transform(list(chain.from_iterable(y_pred)))
tagset = set(lb.classes_) - {'O'}
tagset = sorted(tagset, key=lambda tag: tag.split('-', 1)[::-1])
class_indices = {cls: idx for idx, cls in enumerate(lb.classes_)}
labs = [class_indices[cls] for cls in tagset]
return((precision_recall_fscore_support(y_true_combined,
y_pred_combined,
labels=labs,
average=None,
sample_weight=None)),
(classification_report(
y_true_combined,
y_pred_combined,
labels=[class_indices[cls] for cls in tagset],
target_names=tagset,
)), labs)
def display_image_predictions(features, labels, predictions):
n_classes = 10
label_names = _load_label_names()
label_binarizer = LabelBinarizer()
label_binarizer.fit(range(n_classes))
label_ids = label_binarizer.inverse_transform(np.array(labels))
fig, axies = plt.subplots(nrows=4, ncols=2)
fig.tight_layout()
fig.suptitle('Softmax Predictions', fontsize=20, y=1.1)
n_predictions = 3
margin = 0.05
ind = np.arange(n_predictions)
width = (1. - 2. * margin) / n_predictions
for image_i, (feature, label_id, pred_indicies, pred_values) in enumerate(zip(features, label_ids, predictions.indices, predictions.values)):
pred_names = [label_names[pred_i] for pred_i in pred_indicies]
correct_name = label_names[label_id]
axies[image_i][0].imshow(feature*255)
axies[image_i][0].set_title(correct_name)
axies[image_i][0].set_axis_off()
axies[image_i][1].barh(ind + margin, pred_values[::-1], width)
axies[image_i][1].set_yticks(ind + margin)
axies[image_i][1].set_yticklabels(pred_names[::-1])
axies[image_i][1].set_xticks([0, 0.5, 1.0])
def train_logreg(X, y, test_X, test_y, load_vec=True): """ Trains logistic regression on the feature set. """ full_y = y + test_y lb = LabelBinarizer() lb.fit(full_y) # Convert into 1-D array print len(X), len(test_X) model = LogisticRegression() big_X = X + test_X features = featurize(big_X) X, test_X = features[:4500], features[4500:] print X.shape, X model.fit(X, y) y_pred = model.predict(X) print set(y_pred) print metrics.classification_report(y, y_pred, digits = 3) y_pred = model.predict(test_X) print set(y_pred) print metrics.classification_report(test_y, y_pred, digits = 3)
def Encoding(data, general_matrix=None):
encoder = LabelBinarizer()
count = 0
# encoding
for i in range(data.shape[1]):
if type(data[0, i]) == str:
count += 1
col = data[:, i]
unique = np.unique(col if general_matrix is None else general_matrix[:, i])
try:
encoder.fit(unique)
except:
pass
new_col = encoder.transform(col)
# split at i and i + 1
before, removed, after = np.hsplit(data, [i, i + 1])
# concatenate
data = np.concatenate((before, new_col, after), axis=1)
before, removed, after = np.hsplit(general_matrix, [i, i + 1])
general_matrix = np.concatenate((before, encoder.transform(general_matrix[:, i]), after), axis=1)
print "count : %d" % count
# return data
return data
def binarize_seqfeature(X):
"""
Binarizes the sequence features into 1s and 0s.
Parameters:
===========
- X: (pandas DataFrame) the sequence feature matrix without drug resistance values.
Returns:
========
- binarized: (pandas DataFrame) a binarized sequence feature matrix with columns corresponding to particular amino acids at each position.
- binarizers: (dictionary) a dictionary of binarizer objects for each position.
"""
binarized = pd.DataFrame()
binarizers = dict()
for col in X.columns:
lb = LabelBinarizer()
binarized_cols = lb.fit_transform(X[col])
if len(lb.classes_) == 2:
binarized[col] = pd.Series(binarized_cols[:, 0])
else:
for i, c in enumerate(lb.classes_):
binarized[col + "_" + c] = binarized_cols[:, i]
binarizers[col] = lb
return binarized, binarizers
def bio_classification_report(y_true, y_pred):
lb = LabelBinarizer()
y_true_combined = 1 - lb.fit_transform(list(chain.from_iterable(y_true)))
y_pred_combined = list(chain.from_iterable(y_pred))
tagset = set(lb.classes_) - {'O'}
tagset = sorted(tagset, key=lambda tag: tag.split('-', 1)[::-1])
class_indices = {cls: idx for idx, cls in enumerate(lb.classes_)}
print 'True sum %d Pred sum %d Len %d' %(sum(y_true_combined), sum(y_pred_combined), len(y_pred_combined))
print "AUC\tP-R: %.4f\tROC: %.4f" % (average_precision_score(y_true_combined, y_pred_combined, average=None),
roc_auc_score(y_true_combined, y_pred_combined, average=None))
#plt.figure()
#fpr, tpr, thr = roc_curve(y_true_combined, y_pred_combined)
#area = auc(fpr, tpr)
#plt.plot(fpr, tpr, label='{area:.3f}'.format( area=area))
#plt.legend(loc=4)
#plt.savefig('sub3.jpg')
return classification_report(
1 - y_true_combined,
[0 if v > 0.1 else 1 for v in y_pred_combined],
labels=[class_indices[cls] for cls in tagset],
target_names=tagset,
)
def just_categorical(dropped):
# create initial matrix
print('starting with m0')
lb = LabelBinarizer(sparse_output=True)
m = lb.fit_transform(dropped.restaurant_id)
print(m.shape)
# build matrix
# making nan its own category for categorical
print("adding categorical to matrix")
m = add_categorical_to_matrix(m, dropped, ['review_stars', 'user_name', 'restaurant_stars', 'restaurant_attributes_ages_allowed', 'restaurant_attributes_alcohol', 'restaurant_attributes_attire', 'restaurant_attributes_byob_corkage', 'restaurant_attributes_noise_level', 'restaurant_attributes_smoking', 'restaurant_attributes_wifi', 'restaurant_city', 'restaurant_hours_friday_close', 'restaurant_hours_friday_open', 'restaurant_hours_monday_close', 'restaurant_hours_monday_open', 'restaurant_hours_saturday_close', 'restaurant_hours_saturday_open', 'restaurant_hours_sunday_close', 'restaurant_hours_sunday_open', 'restaurant_hours_thursday_close', 'restaurant_hours_thursday_open', 'restaurant_hours_tuesday_close', 'restaurant_hours_tuesday_open', 'restaurant_hours_wednesday_close', 'restaurant_hours_wednesday_open', 'restaurant_ambience', 'restaurant_music', 'restaurant_parking', 'restaurant_street', 'restaurant_zipcode', 'inspection_year', 'inspection_month', 'inspection_day', 'inspection_dayofweek', 'inspection_quarter',])
print(m.shape)
print("adding bool to matrix")
m = add_categorical_to_matrix(m, dropped, ['restaurant_attributes_accepts_credit_cards', 'restaurant_attributes_byob', 'restaurant_attributes_caters', 'restaurant_attributes_coat_check', 'restaurant_attributes_corkage', 'restaurant_attributes_delivery', 'restaurant_attributes_dietary_restrictions_dairy_free', 'restaurant_attributes_dietary_restrictions_gluten_free', 'restaurant_attributes_dietary_restrictions_halal', 'restaurant_attributes_dietary_restrictions_kosher', 'restaurant_attributes_dietary_restrictions_soy_free', 'restaurant_attributes_dietary_restrictions_vegan', 'restaurant_attributes_dietary_restrictions_vegetarian', 'restaurant_attributes_dogs_allowed', 'restaurant_attributes_drive_thr', 'restaurant_attributes_good_for_dancing', 'restaurant_attributes_good_for_groups', 'restaurant_attributes_good_for_breakfast', 'restaurant_attributes_good_for_brunch', 'restaurant_attributes_good_for_dessert', 'restaurant_attributes_good_for_dinner', 'restaurant_attributes_good_for_latenight', 'restaurant_attributes_good_for_lunch', 'restaurant_attributes_good_for_kids', 'restaurant_attributes_happy_hour', 'restaurant_attributes_has_tv', 'restaurant_attributes_open_24_hours', 'restaurant_attributes_order_at_counter', 'restaurant_attributes_outdoor_seating', 'restaurant_attributes_payment_types_amex', 'restaurant_attributes_payment_types_cash_only', 'restaurant_attributes_payment_types_discover', 'restaurant_attributes_payment_types_mastercard', 'restaurant_attributes_payment_types_visa', 'restaurant_attributes_take_out', 'restaurant_attributes_takes_reservations', 'restaurant_attributes_waiter_service', 'restaurant_attributes_wheelchair_accessible', ])
print(m.shape)
print("adding restaurant categories to matrix")
cats = ['restaurant_category_1', 'restaurant_category_2', 'restaurant_category_3', 'restaurant_category_4', 'restaurant_category_5', 'restaurant_category_6', 'restaurant_category_7']
m = special_categories_to_matrix(m, dropped, cats)
print(m.shape)
print("adding restaurant neighborhoods to matrix")
cats = ['restaurant_neighborhood_1', 'restaurant_neighborhood_2', 'restaurant_neighborhood_3']
m = special_categories_to_matrix(m, dropped, cats)
print(m.shape)
print("matrix shape of {}".format(m.shape))
joblib.dump(m, 'pickle_jar/categorical_matrix')
def transform(self, data_dict):
listOfUnits = ["kilogram", "kg", "gram", "[GMgmkK]?Hz", "liter", "ml",
"cup", "cm", "foot", "inch", "meter", "mg", "gallon", "milliliter", "[MGTmgtKk]B"]
regex = "[\d]+\.[\d]+(" + "[\b/,-]|".join(listOfUnits) + ")"
data = data_dict[self.key].str.extract(regex, flags = re.IGNORECASE, expand=False).str.lower()
lb = LabelBinarizer()
return lb.fit_transform(data.fillna(""))
def log_loss(y_true, y_pred, eps=1e-15, normalize=True, sample_weight=None):
lb = LabelBinarizer()
T = lb.fit_transform(y_true)
if T.shape[1] == 1:
T = np.append(1 - T, T, axis=1)
# Clipping
Y = np.clip(y_pred, eps, 1 - eps)
# This happens in cases when elements in y_pred have type "str".
if not isinstance(Y, np.ndarray):
raise ValueError("y_pred should be an array of floats.")
# If y_pred is of single dimension, assume y_true to be binary
# and then check.
if Y.ndim == 1:
Y = Y[:, np.newaxis]
if Y.shape[1] == 1:
Y = np.append(1 - Y, Y, axis=1)
# Check if dimensions are consistent.
val.check_consistent_length(T, Y)
T = val.check_array(T)
Y = val.check_array(Y)
print(T)
print(Y)
if T.shape[1] != Y.shape[1]:
raise ValueError("y_true and y_pred have different number of classes "
"%d, %d" % (T.shape[1], Y.shape[1]))
# Renormalize
Y /= Y.sum(axis=1)[:, np.newaxis]
loss = -(T * np.log(Y)).sum(axis=1)
return _weighted_sum(loss, sample_weight, normalize)
def full_matrix(dropped):
# create initial matrix
print('starting with m0')
lb = LabelBinarizer(sparse_output=True)
# m = lb.fit_transform(dropped.restaurant_id)
m = lb.fit_transform(dropped.user_name)
print(m.shape)
# build matrix
# making nan its own category for categorical
print("adding categorical to matrix")
m = add_categorical_to_matrix(m, dropped, ['review_stars', 'user_name', 'restaurant_stars', 'restaurant_attributes_ages_allowed', 'restaurant_attributes_alcohol', 'restaurant_attributes_attire', 'restaurant_attributes_byob_corkage', 'restaurant_attributes_noise_level', 'restaurant_attributes_smoking', 'restaurant_attributes_wifi', 'restaurant_city', 'restaurant_hours_friday_close', 'restaurant_hours_friday_open', 'restaurant_hours_monday_close', 'restaurant_hours_monday_open', 'restaurant_hours_saturday_close', 'restaurant_hours_saturday_open', 'restaurant_hours_sunday_close', 'restaurant_hours_sunday_open', 'restaurant_hours_thursday_close', 'restaurant_hours_thursday_open', 'restaurant_hours_tuesday_close', 'restaurant_hours_tuesday_open', 'restaurant_hours_wednesday_close', 'restaurant_hours_wednesday_open', 'restaurant_ambience', 'restaurant_music', 'restaurant_parking', 'restaurant_street', 'restaurant_zipcode', 'inspection_year', 'inspection_month', 'inspection_day', 'inspection_dayofweek', 'inspection_quarter',])
print(m.shape)
print("adding bool to matrix")
m = add_categorical_to_matrix(m, dropped, ['restaurant_attributes_accepts_credit_cards', 'restaurant_attributes_byob', 'restaurant_attributes_caters', 'restaurant_attributes_coat_check', 'restaurant_attributes_corkage', 'restaurant_attributes_delivery', 'restaurant_attributes_dietary_restrictions_dairy_free', 'restaurant_attributes_dietary_restrictions_gluten_free', 'restaurant_attributes_dietary_restrictions_halal', 'restaurant_attributes_dietary_restrictions_kosher', 'restaurant_attributes_dietary_restrictions_soy_free', 'restaurant_attributes_dietary_restrictions_vegan', 'restaurant_attributes_dietary_restrictions_vegetarian', 'restaurant_attributes_dogs_allowed', 'restaurant_attributes_drive_thr', 'restaurant_attributes_good_for_dancing', 'restaurant_attributes_good_for_groups', 'restaurant_attributes_good_for_breakfast', 'restaurant_attributes_good_for_brunch', 'restaurant_attributes_good_for_dessert', 'restaurant_attributes_good_for_dinner', 'restaurant_attributes_good_for_latenight', 'restaurant_attributes_good_for_lunch', 'restaurant_attributes_good_for_kids', 'restaurant_attributes_happy_hour', 'restaurant_attributes_has_tv', 'restaurant_attributes_open_24_hours', 'restaurant_attributes_order_at_counter', 'restaurant_attributes_outdoor_seating', 'restaurant_attributes_payment_types_amex', 'restaurant_attributes_payment_types_cash_only', 'restaurant_attributes_payment_types_discover', 'restaurant_attributes_payment_types_mastercard', 'restaurant_attributes_payment_types_visa', 'restaurant_attributes_take_out', 'restaurant_attributes_takes_reservations', 'restaurant_attributes_waiter_service', 'restaurant_attributes_wheelchair_accessible', ])
print(m.shape)
m = add_numerical_to_matrix(m, dropped, ['review_votes_cool', 'review_votes_funny', 'review_votes_useful', 'user_average_stars', 'user_compliments_cool', 'user_compliments_cute', 'user_compliments_funny', 'user_compliments_hot', 'user_compliments_list', 'user_compliments_more', 'user_compliments_note', 'user_compliments_photos', 'user_compliments_plain', 'user_compliments_profile', 'user_compliments_writer', 'user_fans', 'user_review_count', 'user_votes_cool', 'user_votes_funny', 'user_votes_useful', 'restaurant_attributes_price_range', 'restaurant_latitude', 'restaurant_longitude', 'restaurant_review_count', 'checkin_counts', 'review_delta', 'previous_inspection_delta', 'polarity', 'subjectivity', 'neg', 'neu', 'pos', 'compound', 'user_yelping_since_delta','manager', 'supervisor', 'training', 'safety', 'disease', 'ill', 'sick', 'poisoning', 'hygiene', 'raw', 'undercooked', 'cold', 'clean', 'sanitary', 'wash', 'jaundice', 'yellow', 'hazard', 'inspection', 'violation', 'gloves', 'hairnet', 'nails', 'jewelry', 'sneeze', 'cough', 'runny', 'illegal', 'rotten', 'dirty', 'mouse', 'cockroach', 'contaminated', 'gross', 'disgusting', 'stink', 'old', 'parasite', 'reheat', 'frozen', 'broken', 'drip', 'bathroom', 'toilet', 'leak', 'trash', 'dark', 'lights', 'dust', 'puddle', 'pesticide', 'bugs', 'mold'])
print(m.shape)
print("adding restaurant categories to matrix")
cats = ['restaurant_category_1', 'restaurant_category_2', 'restaurant_category_3', 'restaurant_category_4', 'restaurant_category_5', 'restaurant_category_6', 'restaurant_category_7']
m = special_categories_to_matrix(m, dropped, cats)
print(m.shape)
print("adding restaurant neighborhoods to matrix")
cats = ['restaurant_neighborhood_1', 'restaurant_neighborhood_2', 'restaurant_neighborhood_3']
m = special_categories_to_matrix(m, dropped, cats)
print(m.shape)
print("matrix shape of {}".format(m.shape))
joblib.dump(m, 'pickle_jar/full_matrix')
class BusinessCategoriesFeature(BaseEstimator): """ WARNING!!! Works only with a modified version of LabelBinarizer. A binarization of the reviews' business categories. """ def __init__(self, data=None): self.data = data def __create_labels_list(self, review_list): labels = [] for review in review_list: business = self.data.get_business_for_review(review) labels.append(business['categories']) return labels def fit(self, X, y): self.binarizer = LabelBinarizer() labels = self.__create_labels_list(X) self.binarizer.fit(labels) return self def transform(self, X): labels = self.__create_labels_list(X) binarized_labels = self.binarizer.transform(labels) return binarized_labels.astype(float)
def fit(self, X, y):
"""
performs one step of gradient descent
"""
# get the dimensions of our data
n_samples, n_features = X.shape[0], X.shape[1]+1
n_targets = len(np.unique(y))
# add a column to the data matrix to incorporate the bias term
X = np.c_[np.ones(n_samples), X]
# one-vs-all labeling
lb = LabelBinarizer()
y = lb.fit_transform(y)
# initialize the weights
if self.W is None:
self.W = np.zeros( (n_features, n_targets) )
# perform the optimization using gradient descent with momentum
grad = self.gradient(X,y)
self.W = self.W - self.learning_rate*(grad + self.momentum*self.prev_grad)
self.prev_grad = grad
return self.loss(X,y)
def run():
# Load and preprocess data
label_to_unique_instance = load_data()
X, Y = preprocess_data(label_to_unique_instance)
# Encode labels
label_binarizer = LabelBinarizer()
transformed_Y = label_binarizer.fit_transform(Y)
# Cross validation
cross_validation_iterator = StratifiedShuffleSplit(Y, n_iter=1, test_size=0.4, random_state=0)
for train_index, test_index in cross_validation_iterator:
break
# Init model
model = init_model(raw_feature_dim=X.shape[-1], unique_lable_num=len(label_binarizer.classes_))
# Training procedure
model.fit(X[train_index], transformed_Y[train_index],
batch_size=BATCH_SIZE, nb_epoch=MAXIMUM_EPOCH_NUM,
validation_data=(X[test_index], transformed_Y[test_index]),
callbacks=[TensorBoard(log_dir="/tmp/Sequence Classification")],
verbose=2)
print("All done!")
def report(test_y, pred_y):
lb = LabelBinarizer()
test_y_combined = lb.fit_transform(list(chain.from_iterable(test_y)))
pred_y_combined = lb.transform(list(chain.from_iterable(pred_y)))
tagset = sorted(set(lb.classes_))
class_indices = {cls: idx for idx, cls in enumerate(tagset)}
print(classification_report(test_y_combined, pred_y_combined, labels=[class_indices[cls] for cls in tagset], target_names=tagset))
def conv_demo():
# load the digits dataset
digits = load_digits()
X = digits['data']
y_labels = digits['target']
lb = LabelBinarizer()
y = lb.fit_transform(y_labels)
# split into training, validation and test datasets
X_train, X_test, y_train, y_test = train_test_split(X, y,
test_size=0.25,
random_state=RANDOM_STATE)
X_train, X_valid, y_train, y_valid = train_test_split(X_train, y_train,
test_size=0.25,
random_state=RANDOM_STATE)
# train the neural net
print("Building neural net to classify digits")
conv_net = pynn.ConvNet(digits['images'][0].shape, 1, y.shape[1],
random_state=RANDOM_STATE)
print("Training")
conv_net.fit(X_train, y_train, X_valid, y_valid,
batch_size=20, n_epochs=20, learning_rate=0.05)
y_pred = conv_net.predict(X_test)
print("digits accuracy: {}%".format(
accuracy_score(y_test.argmax(1), y_pred.argmax(1)) * 100))
class CategoricalToNumerical(object):
def __init__(self, dimensionality_reducer=None, verify=True):
pass
"""Takes in a dimensionality reducer in order to convert categorical features into numerical.
"""
if dimensionality_reducer is None:
dimensionality_reducer = RandomizedPCA(1)
self.dimensionality_reducer = dimensionality_reducer
self.verify = verify
self.binarizer = LabelBinarizer()
def fit(self, X, y=None):
self._verify(X, self.verify)
binarized = self.binarizer.fit_transform(X)
self.dimensionality_reducer.fit(binarized)
def transform(self, X):
self._verify(X, False)
binarized = self.binarizer.transform(X)
result = self.dimensionality_reducer.transform(binarized).flatten()
assert X.shape == result.shape
return result
def fit_transform(self, X, y=None):
self.fit(X)
return self.transform(X)
def _verify(self, X, verify):
if verify:
assert is_categorical(X)
else:
assert isinstance(X, np.ndarray)
assert len(X.shape) == 1
def get_dataset2(test_fraction):
"""
@:param: test_fraction used to split train and test
Vectorizes the features and labels into categorical values and randomly splits into train and test set
:return: X_train, X_test, y_train, y_test
"""
data = []
with open('labels.csv', 'r') as datafile:
csv_reader = csv.reader(datafile, delimiter=',', quotechar='|')
for row in csv_reader:
data.append(row)
data = numpy.asarray(data)
X = data[:, 0:data.shape[1]-1]
y = data[:, data.shape[1]-1]
# X,y = get_tabledata()
vec = DictVectorizer()
feature_dict = [dict(enumerate(x)) for x in X.tolist()]
X = vec.fit_transform(feature_dict).toarray()
joblib.dump(vec, 'vectorizer.pkl')
lb = LabelBinarizer()
y = lb.fit_transform(y)
joblib.dump(lb, 'binarizer.pkl')
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=test_fraction)
return X_train, X_test, y_train, y_test
def bio_classification_report(y_true, y_pred):
"""Evaluates entity extraction accuracy.
Classification report for a list of BIO-encoded sequences.
It computes token-level metrics and discards "O" labels.
Note that it requires scikit-learn 0.15+ (or a version from github master)
to calculate averages properly!
Taken from https://github.com/scrapinghub/python-crfsuite/blob/master/examples/CoNLL%202002.ipynb
"""
from sklearn.preprocessing import LabelBinarizer
from itertools import chain
from sklearn.metrics import classification_report
lb = LabelBinarizer()
y_true_combined = lb.fit_transform(list(chain.from_iterable(y_true)))
y_pred_combined = lb.transform(list(chain.from_iterable(y_pred)))
tagset = set(lb.classes_) - {'O'}
tagset = sorted(tagset, key=lambda tag: tag.split('-', 1)[::-1])
class_indices = {cls: idx for idx, cls in enumerate(lb.classes_)}
return classification_report(
y_true_combined,
y_pred_combined,
labels=[class_indices[cls] for cls in tagset],
target_names=tagset,
)
def binarize_label_columns(df, columns, two_classes_as='single'):
'''
Inputs:
df: Pandas dataframe object.
columns: Columns to binarize.
tow_classes_as: How to handle two classes, as 'single' or 'multiple' columns.
Returns a tuple with the following items:
df: Pandas dataframe object with new columns.
binlabel_names: Names of the newly created binary variables.
lb_objects: a dictionary with columns as keys and sklear.LabelBinarizer
objects as values.
'''
binlabel_names = []
lb_objects = {}
for col in columns:
if len(df[col].unique()) > 1:
rows_notnull = df[col].notnull() # Use only valid feature observations
lb = LabelBinarizer()
binclass = lb.fit_transform(df[col][rows_notnull]) # Fit & transform on valid observations
if len(lb.classes_) == 2 and two_classes_as == 'multiple':
binclass = np.hstack((1 - binclass, binclass))
lb_objects[col] = lb
if len(lb.classes_) > 2 or two_classes_as == 'multiple':
col_binlabel_names = [col+'_'+str(c) for c in lb.classes_]
binlabel_names += col_binlabel_names # Names for the binarized classes
for n in col_binlabel_names: df[n] = np.NaN # Initialize columns
df.loc[rows_notnull, col_binlabel_names] = binclass # Merge binarized data
elif two_classes_as == 'single':
binlabel_names.append(col+'_bin') # Names for the binarized classes
df[col+'_bin'] = np.NaN # Initialize columns
df.loc[rows_notnull, col+'_bin'] = binclass # Merge binarized data
return df, binlabel_names, lb_objects
def scorer_auc(y_true, y_pred):
from sklearn.metrics import roc_auc_score
from sklearn.preprocessing import LabelBinarizer
"""Dedicated to 2class probabilistic outputs"""
le = LabelBinarizer()
y_true = le.fit_transform(y_true)
return roc_auc_score(y_true, y_pred)
def iris_demo():
# load the iris dataset
iris = load_iris()
X = iris['data']
y_labels = iris['target']
lb = LabelBinarizer()
y = lb.fit_transform(y_labels)
# split into training, validation and test datasets
X_train, X_test, y_train, y_test = train_test_split(X, y,
test_size=0.25,
random_state=RANDOM_STATE)
X_train, X_valid, y_train, y_valid = train_test_split(X_train, y_train,
test_size=0.25,
random_state=RANDOM_STATE)
# train the neural net
print("Building logistic regression classifier to classify iris data")
nn = pynn.ArtificialNeuralNet([X_train.shape[1], 20, y_train.shape[1]])
print("Training")
nn.fit(X_train, y_train, X_valid, y_valid,
batch_size=20, n_epochs=20, learning_rate=0.05,
random_state=RANDOM_STATE)
y_pred = nn.predict(X_test)
print("iris accuracy: {}%".format(
accuracy_score(y_test.argmax(1), y_pred.argmax(1)) * 100))
def one_hot_encoding(y_train, y_test):
labelBinarizer = LabelBinarizer()
labelBinarizer.fit(y_train)
y_train_one_hot = labelBinarizer.transform(y_train)
y_test_one_hot = labelBinarizer.transform(y_test)
return y_train_one_hot, y_test_one_hot
class BaseSGD(object):
def _get_loss(self):
losses = {
"modified_huber": ModifiedHuber(),
"hinge": Hinge(1.0),
"perceptron": Hinge(0.0),
"log": Log(),
"sparse_log": SparseLog(),
"squared": SquaredLoss(),
"huber": Huber(self.epsilon),
"epsilon_insensitive": EpsilonInsensitive(self.epsilon),
}
return losses[self.loss]
def _get_learning_rate(self):
learning_rates = {"constant": 1, "pegasos": 2, "invscaling": 3}
return learning_rates[self.learning_rate]
def _set_label_transformers(self, y):
if self.multiclass == "natural":
self.label_encoder_ = LabelEncoder()
y = self.label_encoder_.fit_transform(y).astype(np.float64)
self.label_binarizer_ = LabelBinarizer(neg_label=-1, pos_label=1)
self.label_binarizer_.fit(y)
self.classes_ = self.label_binarizer_.classes_.astype(np.int32)
n_classes = len(self.label_binarizer_.classes_)
n_vectors = 1 if n_classes <= 2 else n_classes
return n_classes, n_vectors
import argparse
# construct the argument parser and parse the arguments
ap = argparse.ArgumentParser()
ap.add_argument("-o", "--output", required=True,
help="path to the output loss/accuracy plot")
args = vars(ap.parse_args())
# load the training and testing data, then scale it into the range [0, 1]
print("[INFO] loading the CIFAR-10 data...")
((trainX, trainY), (testX, testY)) = cifar10.load_data()
trainX = trainX.astype("float")/255.0
testX = testX.astype("float")/255.0
# convert the labels from integers to vectors
lb = LabelBinarizer()
trainY = lb.fit_transform(trainY)
testY = lb.fit_transform(testY)
# initialize the label names for CIFAR-10 dataset
labelNames=["airplane", "automobile", "bird", "cat", "deer",
"dog", "frog", "horse", "ship", "truck"]
# initialize the optimizer and model
print("[INFO] compiling the model...")
opt = SGD(lr=0.001, decay=0.01/40, momentum=0.9, nesterov=True)
model = MiniVGGNet.build(width=32, height=32, depth=3, classes=10)
model.compile(loss="categorical_crossentropy", optimizer=opt,
metrics=["accuracy"])
print(model.summary())
print(data.head())
print ("\nFeatures : \n" ,data.columns.tolist())
print ("\nMissing values : ", data.isnull().sum().values.sum())
print ("\nUnique values : \n",data.nunique())
for i in data.columns:
data[i] = data[i].replace("?",np.nan)
data = data[data[i].notnull()]
data = data.reset_index()[data.columns]
data[i] = data[i].astype(float)
X = data.drop(['income'], axis=1)
y = data['income']
from sklearn.preprocessing import LabelBinarizer
encoder = LabelBinarizer()
Y = encoder.fit_transform(y)
from keras.models import Sequential #Sequential Models
from keras.layers import Dense #Dense Fully Connected Layer Type
from keras.optimizers import SGD #Stochastic Gradient Descent Optimizer
def create_network():
model = Sequential()
model.add(Dense(25, input_shape=(13,), activation='relu'))
model.add(Dense(9, activation='softmax'))
#stochastic gradient descent
image = img_to_array(image)
data.append(image)
# extract the class label from the image path and update the label list
label = imagePath.split(os.path.sep)[-2]
#print(str(imagePath)+"STR"+str(label))
labels.append(label)
# scale the raw pixel intensities to the range [0, 1]
data = np.array(data, dtype="float") / 255.0
labels = np.array(labels)
#print(labels)
print("[INFO] data matrix: {:.2f}MB".format(data.nbytes / (1024 * 1000.0)))
# binarize the labels
lb = LabelBinarizer()
labels = lb.fit_transform(labels)
print("LB.CLASSES=" + str(lb.classes_))
# partition data into train and validation
(trainX, testX, trainY, testY) = train_test_split(data,
labels,
test_size=0.2,
random_state=42)
# construct image generator for data augmentation
aug = ImageDataGenerator(rotation_range=25,
width_shift_range=0.1,
height_shift_range=0.1,
shear_range=0.2,
zoom_range=0.2,
dataUni = stretch_interp(data, dur)
dataUni = np.array(dataUni, dtype="float")
(train_x, test_x, train_y, test_y) = train_test_split(dataUni,
labels,
test_size=0.1,
random_state=14)
#train_x = train_x[:, :, np.newaxis]
#test_x = test_x[:, :, np.newaxis]
print(np.shape(train_x), np.shape(train_y), np.shape(test_x), np.shape(test_y))
print("labels:", train_y)
lb = LabelBinarizer()
train_y = lb.fit_transform(train_y)
test_y = lb.transform(test_y)
print("labels after transform:", train_y)
print(np.shape(train_x), np.shape(train_y), np.shape(test_x), np.shape(test_y))
batch_size = np.shape(train_x)[0]
model = Sequential()
#model.add(Conv1D(3012, input_shape=(53, dur*3012), kernel_size=(12), activation='relu'))
model.add(Dense(3012, activation="relu", input_shape=(dur * 3012, )))
print("after 1 dense:", model.input_shape, model.output_shape)
model.add(Dense(506, activation="sigmoid"))
model.add(Dropout(0.25))
model.add(Dense(253, activation="sigmoid"))
def train_on_texts(self, texts, context_labels=None,
batch_size=128,
num_epochs=50,
verbose=1,
new_model=False,
gen_epochs=1,
train_size=1.0,
max_gen_length=300,
validation=True,
dropout=0.0,
via_new_model=False,
save_epochs=0,
multi_gpu=False,
**kwargs):
if new_model and not via_new_model:
self.train_new_model(texts,
context_labels=context_labels,
num_epochs=num_epochs,
gen_epochs=gen_epochs,
train_size=train_size,
batch_size=batch_size,
dropout=dropout,
validation=validation,
save_epochs=save_epochs,
multi_gpu=multi_gpu,
**kwargs)
return
if context_labels:
context_labels = LabelBinarizer().fit_transform(context_labels)
if 'prop_keep' in kwargs:
train_size = prop_keep
if self.config['word_level']:
# If training word level, must add spaces around each
# punctuation. https://stackoverflow.com/a/3645946/9314418
punct = '!"#$%&()*+,-./:;<=>?@[\]^_`{|}~\\n\\t\'‘’“”’–—…'
for i in range(len(texts)):
texts[i] = re.sub('([{}])'.format(punct), r' \1 ', texts[i])
texts[i] = re.sub(' {2,}', ' ', texts[i])
texts = [text_to_word_sequence(text, filters='') for text in texts]
# calculate all combinations of text indices + token indices
indices_list = [np.meshgrid(np.array(i), np.arange(
len(text) + 1)) for i, text in enumerate(texts)]
# indices_list = np.block(indices_list) # this hangs when indices_list is large enough
# FIX BEGIN ------
indices_list_o = np.block(indices_list[0])
for i in range(len(indices_list)-1):
tmp = np.block(indices_list[i+1])
indices_list_o = np.concatenate([indices_list_o, tmp])
indices_list = indices_list_o
# FIX END ------
# If a single text, there will be 2 extra indices, so remove them
# Also remove first sequences which use padding
if self.config['single_text']:
indices_list = indices_list[self.config['max_length']:-2, :]
indices_mask = np.random.rand(indices_list.shape[0]) < train_size
if multi_gpu:
num_gpus = len(config.get_visible_devices('GPU'))
batch_size = batch_size * num_gpus
gen_val = None
val_steps = None
if train_size < 1.0 and validation:
indices_list_val = indices_list[~indices_mask, :]
gen_val = generate_sequences_from_texts(
texts, indices_list_val, self, context_labels, batch_size)
val_steps = max(
int(np.floor(indices_list_val.shape[0] / batch_size)), 1)
indices_list = indices_list[indices_mask, :]
num_tokens = indices_list.shape[0]
assert num_tokens >= batch_size, "Fewer tokens than batch_size."
level = 'word' if self.config['word_level'] else 'character'
print("Training on {:,} {} sequences.".format(num_tokens, level))
steps_per_epoch = max(int(np.floor(num_tokens / batch_size)), 1)
gen = generate_sequences_from_texts(
texts, indices_list, self, context_labels, batch_size)
base_lr = 4e-3
# scheduler function must be defined inline.
def lr_linear_decay(epoch):
return (base_lr * (1 - (epoch / num_epochs)))
'''
FIXME
This part is a bit messy as we need to initialize the model within
strategy.scope() when using multi-GPU. Can probably be cleaned up a bit.
'''
if context_labels is not None:
if new_model:
weights_path = None
else:
weights_path = "{}_weights.hdf5".format(self.config['name'])
self.save(weights_path)
if multi_gpu:
from tensorflow import distribute as distribute
strategy = distribute.MirroredStrategy()
with strategy.scope():
parallel_model = textgenrnn_model(self.num_classes,
dropout=dropout,
cfg=self.config,
context_size=context_labels.shape[1],
weights_path=weights_path)
parallel_model.compile(loss='categorical_crossentropy',
optimizer=Adam(lr=4e-3))
model_t = parallel_model
print("Training on {} GPUs.".format(num_gpus))
else:
model_t = self.model
else:
if multi_gpu:
from tensorflow import distribute as distribute
if new_model:
weights_path = None
else:
weights_path = "{}_weights.hdf5".format(self.config['name'])
strategy = distribute.MirroredStrategy()
with strategy.scope():
# Do not locate model/merge on CPU since sample sizes are small.
parallel_model = textgenrnn_model(self.num_classes,
cfg=self.config,
weights_path=weights_path)
parallel_model.compile(loss='categorical_crossentropy',
optimizer=Adam(lr=4e-3))
model_t = parallel_model
print("Training on {} GPUs.".format(num_gpus))
else:
model_t = self.model
model_t.fit(gen, steps_per_epoch=steps_per_epoch,
epochs=num_epochs,
callbacks=[
LearningRateScheduler(
lr_linear_decay),
generate_after_epoch(
self, gen_epochs,
max_gen_length),
save_model_weights(
self, num_epochs,
save_epochs)],
verbose=verbose,
max_queue_size=10,
validation_data=gen_val,
validation_steps=val_steps
)
# Keep the text-only version of the model if using context labels
if context_labels is not None:
self.model = Model(inputs=self.model.input[0],
outputs=self.model.output[1])
def fit(self, X, y):
self.label_binarizer_ = LabelBinarizer(neg_label=-1, pos_label=1)
Y = np.asfortranarray(self.label_binarizer_.fit_transform(y),
dtype=np.float64)
return self._fit(X, Y)
from NeuralNetwork import NeuralNetwork
from sklearn.preprocessing import LabelBinarizer
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report
from sklearn import datasets
print("[INFO] loading MNIST (sample) dataset...")
digits = datasets.load_digits()
data = digits.data.astype("float")
data = (data - data.min()) / (data.max() - data.min())
print("[INFO] samples: {}, dim: {}".format(data.shape[0], data.shape[1]))
(trainX, testX, trainY, testY) = train_test_split(data,
digits.target,
test_size=0.25)
trainY = LabelBinarizer().fit_transform(trainY)
testY = LabelBinarizer().fit_transform(testY)
print("[INFO] trainig network...")
nn = NeuralNetwork([trainX.shape[1], 32, 16, 10])
print("[INFO] {}".format(nn))
nn.fit(trainX, trainY, epochs=1000)
print("[INFO] evaluating network...")
prediction = nn.predict(testX)
prediction = predictions.argmax(axis=1)
print(classification_report(testY.argmax(axis=1), predictions))
def fit(self, X, y, sample_weight=None):
"""Fit the model according to the given training data.
Parameters
----------
X : {array-like, sparse matrix}, shape (n_samples, n_features)
Training vector, where n_samples is the number of samples and
n_features is the number of features.
y : array-like, shape (n_samples,)
Target vector relative to X.
sample_weight : array-like, shape (n_samples,) optional
Array of weights that are assigned to individual samples.
If not provided, then each sample is given unit weight.
.. versionadded:: 0.17
*sample_weight* support to LogisticRegression.
Returns
-------
self : object
Returns self.
"""
self._label_binarizer = LabelBinarizer(pos_label=1, neg_label=-1)
Y = [self._label_binarizer.fit_transform(yy).ravel() for yy in y]
# self.lr_p2 = [SGDClassifier(
# loss='log', l1_ratio=self.l1_ratio_lamda,
# fit_intercept=self.fit_intercept, shuffle=False,
# penalty='elasticnet', alpha=self.lamda, warm_start=True,
# max_iter=(self.max_iter // 3 if self.deep else 1) + 0)
# for i in range(len(X))]
self.lr_p2 = [
LogisticRegression(
fit_intercept=self.fit_intercept,
penalty='l2',
solver='lbfgs',
C=1. / (self.lamda * (1 - self.l1_ratio_lamda)),
warm_start=True,
max_iter=(self.max_iter // 3 if self.deep else 1) + 5)
for i in range(len(X))
]
self.alpha_, self.coef_, self.intercept_, self.n_iter_ = \
logistic_alternating(
X, Y, lamda=self.lamda, beta=self.beta, gamma=self.gamma,
max_iter=self.max_iter, verbose=self.verbose, tol=self.tol,
return_n_iter=True, deep=self.deep,
lr_p2=self.lr_p2, l1_ratio_beta=self.l1_ratio_beta,
l1_ratio_lamda=self.l1_ratio_lamda, # unused
fit_intercept=self.fit_intercept # unused
)
if self.classes_.shape[0] > 2:
# ndim = self.classes_.shape[0]
raise ValueError("too many classes")
else:
ndim = 1
self.coef_ = self.coef_.reshape(ndim, -1)
# self.alpha_ = [alpha.reshape(ndim, -1) for alpha in self.alpha_]
self.y_train_ = Y
return self
class LocalTransformer(BaseEstimator, TransformerMixin):
def __init__(self,
k=1,
fxs=None,
hour_mean=True,
aux=True,
ens_std=False,
hour_std=False,
stid_enc=False):
self.k = k
self.hour_mean = hour_mean
self.fxs = fxs
self.aux = aux
self.ens_std = ens_std
self.hour_std = hour_std
self.stid_enc = stid_enc
def fit(self, X, y=None):
assert y is not None
n_stations = y.shape[1]
self.n_stations = n_stations
assert X.station_info.shape[0] == self.n_stations
stid = np.arange(98)
self.stid_lb = LabelBinarizer()
self.stid_lb.fit(stid)
return self
@classmethod
def transform_labels(cls, y):
y = y.ravel(1)
return y
# @profile
def transform(self, X, y=None):
k = self.k
n_days = X.shape[0]
ll_coord = X.station_info[:, 0:2]
lat_idx = np.searchsorted(X.lat, ll_coord[:, 0])
lon_idx = np.searchsorted(X.lon, ll_coord[:, 1] + 360)
#IPython.embed()
n_fx = 0
for b_name, X_b in X.blocks.iteritems():
old_n_fx = n_fx
if self.fxs is not None and b_name not in self.fxs:
continue
if X_b.ndim == 6:
if self.fxs is not None and b_name in self.fxs:
n_fxs = len(self.fxs[b_name])
else:
n_fxs = X_b.shape[1]
shapes = [n_fxs]
if not self.hour_mean:
shapes.append(X_b.shape[3])
shapes.extend([k * 2, k * 2])
print b_name, shapes
n_fx += np.prod(shapes)
elif X_b.ndim == 1:
n_fx += 1
elif X_b.ndim == 2:
n_fx += X_b.shape[1]
else:
raise ValueError('%s has wrong dim: %d' % (b_name, X_b.ndim))
print 'block: %s as %d n_fx' % (b_name, n_fx - old_n_fx)
if self.stid_enc:
n_fx += len(self.stid_lb.classes_)
# num of features - based on blocks + station info (5 fx)
if self.aux:
n_fx = n_fx + 3 + 2 + 2
X_p = np.zeros((n_days * self.n_stations, n_fx), dtype=np.float32)
offset = 0
for bid, b_name in enumerate(X.blocks):
if self.fxs is not None and b_name not in self.fxs:
continue
print 'localizing block: %s' % b_name
X_b = X[b_name]
# select fx if fxs given
if self.fxs is not None and self.fxs.get(b_name, None):
fxs = self.fxs[b_name]
idx = [
i for i, name in enumerate(X.fx_name[b_name])
if name in fxs
]
X_b = X_b[:, idx]
if X_b.ndim == 6:
# FIXME over hours
if self.hour_mean:
X_b = np.mean(X_b, axis=3)
elif self.hour_std:
X_b = np.std(X_b, axis=3)
# over ensembles
if self.ens_std:
X_b = np.std(X_b, axis=2)
else:
X_b = np.mean(X_b, axis=2)
offset_inc = 0
for i in range(self.n_stations):
lai, loi = lat_idx[i], lon_idx[i]
if (self.hour_mean or self.hour_std):
blk = X_b[:, :, lai - k:lai + k, loi - k:loi + k]
else:
blk = X_b[:, :, :, lai - k:lai + k, loi - k:loi + k]
blk = blk.reshape((blk.shape[0], np.prod(blk.shape[1:])))
X_p[i * n_days:((i + 1) * n_days),
offset:(offset + blk.shape[1])] = blk
if i == 0:
offset_inc = blk.shape[1]
del blk
gc.collect()
offset += offset_inc
elif X_b.ndim == 1 or (X_b.ndim == 2 and X_b.shape[1] == 1):
X_p[:,
offset:offset + 1] = np.tile(X_b.ravel(),
self.n_stations)[:,
np.newaxis]
offset += 1
elif X_b.ndim == 2:
# FIXME wrong stitching together stuff
print('block: %s will be repeated for each station' % b_name)
width = X_b.shape[1]
X_p[:,
offset:offset + width] = np.tile(X_b, (self.n_stations, 1))
#IPython.embed()
offset += width
else:
raise ValueError('%s has wrong dim: %d' % (b_name, X_b.ndim))
if self.stid_enc:
stid = np.repeat(self.stid_lb.classes_, n_days)
stid_enc = self.stid_lb.transform(stid)
X_p[:, offset:(offset + stid_enc.shape[1])] = stid_enc
offset += stid_enc.shape[1]
if self.aux:
# lat, lon, elev
X_p[:, offset:(offset + 3)] = np.repeat(X.station_info,
n_days,
axis=0)
offset += 3
# compute pos of station within grid cell (in degree lat lon)
lat_idx = np.repeat(lat_idx, n_days)
lon_idx = np.repeat(lon_idx, n_days)
# offset - 3 is station lat
X_p[:, offset] = (X_p[:, offset - 3] - X.lat[lat_idx])
# offset - 2 is station lon
X_p[:, offset + 1] = (X_p[:, offset - 2] - (X.lon[lon_idx] - 360.))
# FIXME add lat lon idx
offset += 2
X_p[:, offset] = lat_idx
X_p[:, offset + 1] = lon_idx
print 'X_p.shape: ', X_p.shape
return X_p
import numpy as np
from NeuralNet import NeuralNet
from sklearn.preprocessing import LabelBinarizer
train = np.loadtxt('./wine/train_wine.csv', delimiter=',')
lb = LabelBinarizer()
train_y = lb.fit_transform(train[:, 0])
train_x = train[:, 1:]
mean = np.mean(train_x, axis=0)
train_x -= mean
var = np.var(train_x, axis=0)
train_x /= var
nn = NeuralNet([13, 13, 3], 0.01)
nn.train(train_x, train_y)
test = np.loadtxt('./wine/test_wine.csv', delimiter=',')
test_y = test[:, 0]
test_x = test[:, 1:]
test_x -= mean
test_x /= var
y_pred = np.argmax(nn.predict(test_x), axis=1)
accu = np.sum(test_y == y_pred + 1) / len(test_y)
print(accu)
dpt = 3
print("[INFO] loading Images")
imagePaths = list(paths.list_images(args["dataset"]))
sp = SimplePreprocessor(size, size)
iap = ImageToArrayPreprocessor()
sdl = SimpleDatasetLoader(preprocessors=[sp, iap])
(data, labels) = sdl.load(imagePaths, verbose=500)
print(labels)
data = data.astype("float") / 255.0
(trainX, testX, trainY, testY) = train_test_split(data,
labels,
test_size=0.25,
random_state=42)
trainY = LabelBinarizer().fit_transform(trainY)
testY = LabelBinarizer().fit_transform(testY)
print("[INFO] compiling model...")
opt = SGD(lr=0.025)
model = IncludeNet.build(width=size, height=size, depth=dpt, classes=4)
model.compile(loss="categorical_crossentropy",
optimizer=opt,
metrics=["accuracy"])
print("[INFO] training network...")
H = model.fit(trainX,
trainY,
validation_data=(testX, testY),
batch_size=size,
epochs=ep,
class MultipleLogisticRegressionMultipleKernel(
LogisticRegressionMultipleKernel, LogisticRegression,
LinearClassifierMixin):
# Ensure consistent split
_pairwise = True
def __init__(self,
penalty='l2',
dual=False,
tol=1e-4,
fit_intercept=True,
intercept_scaling=1,
class_weight=None,
random_state=None,
solver='liblinear',
max_iter=100,
multi_class='ovr',
verbose=0,
warm_start=False,
n_jobs=1,
l1_ratio_lamda=0.1,
l1_ratio_beta=0.1,
deep=True,
lamda=0.01,
gamma=1,
rho=1,
rtol=1e-4,
beta=0.01):
super(MultipleLogisticRegressionMultipleKernel,
self).__init__(penalty=penalty,
dual=dual,
tol=tol,
fit_intercept=fit_intercept,
intercept_scaling=intercept_scaling,
class_weight=class_weight,
random_state=random_state,
solver=solver,
max_iter=max_iter,
multi_class=multi_class,
verbose=verbose,
warm_start=warm_start,
n_jobs=n_jobs,
lamda=lamda,
gamma=gamma,
rho=rho,
rtol=rtol,
beta=beta)
self.l1_ratio_lamda = l1_ratio_lamda
self.l1_ratio_beta = l1_ratio_beta
self.deep = deep
def fit(self, X, y, sample_weight=None):
"""Fit the model according to the given training data.
Parameters
----------
X : {array-like, sparse matrix}, shape (n_samples, n_features)
Training vector, where n_samples is the number of samples and
n_features is the number of features.
y : array-like, shape (n_samples,)
Target vector relative to X.
sample_weight : array-like, shape (n_samples,) optional
Array of weights that are assigned to individual samples.
If not provided, then each sample is given unit weight.
.. versionadded:: 0.17
*sample_weight* support to LogisticRegression.
Returns
-------
self : object
Returns self.
"""
self._label_binarizer = LabelBinarizer(pos_label=1, neg_label=-1)
Y = [self._label_binarizer.fit_transform(yy).ravel() for yy in y]
# self.lr_p2 = [SGDClassifier(
# loss='log', l1_ratio=self.l1_ratio_lamda,
# fit_intercept=self.fit_intercept, shuffle=False,
# penalty='elasticnet', alpha=self.lamda, warm_start=True,
# max_iter=(self.max_iter // 3 if self.deep else 1) + 0)
# for i in range(len(X))]
self.lr_p2 = [
LogisticRegression(
fit_intercept=self.fit_intercept,
penalty='l2',
solver='lbfgs',
C=1. / (self.lamda * (1 - self.l1_ratio_lamda)),
warm_start=True,
max_iter=(self.max_iter // 3 if self.deep else 1) + 5)
for i in range(len(X))
]
self.alpha_, self.coef_, self.intercept_, self.n_iter_ = \
logistic_alternating(
X, Y, lamda=self.lamda, beta=self.beta, gamma=self.gamma,
max_iter=self.max_iter, verbose=self.verbose, tol=self.tol,
return_n_iter=True, deep=self.deep,
lr_p2=self.lr_p2, l1_ratio_beta=self.l1_ratio_beta,
l1_ratio_lamda=self.l1_ratio_lamda, # unused
fit_intercept=self.fit_intercept # unused
)
if self.classes_.shape[0] > 2:
# ndim = self.classes_.shape[0]
raise ValueError("too many classes")
else:
ndim = 1
self.coef_ = self.coef_.reshape(ndim, -1)
# self.alpha_ = [alpha.reshape(ndim, -1) for alpha in self.alpha_]
self.y_train_ = Y
return self
def predict(self, X):
"""Predict using the kernel ridge model.
Parameters
----------
X : {array-like, sparse matrix}, shape = [n_samples, n_features]
Samples.
Returns
-------
C : array, shape = [n_samples] or [n_samples, n_targets]
Returns predicted values.
"""
check_is_fitted(self, ["alpha_", "coef_"])
# return [LinearClassifierMixin.predict(
# self, np.tensordot(k, a, axes=1)) for a, k in zip(
# self.alpha_, X)]
return [
self.lr_p2[i].predict(
np.tensordot(self.coef_.ravel(), X[i], axes=1))
for i in range(len(X))
]
def score(self, K, y, sample_weight=None):
"""Returns the coefficient of determination R^2 of the prediction.
The coefficient R^2 is defined as (1 - u/v), where u is the residual
sum of squares ((y_true - y_pred) ** 2).sum() and v is the total
sum of squares ((y_true - y_true.mean()) ** 2).sum().
The best possible score is 1.0 and it can be negative (because the
model can be arbitrarily worse). A constant model that always
predicts the expected value of y, disregarding the input features,
would get a R^2 score of 0.0.
Parameters
----------
X : array-like, shape = (n_samples, n_features)
Test samples.
y : array-like, shape = (n_samples) or (n_samples, n_outputs)
True values for X.
sample_weight : array-like, shape = [n_samples], optional
Sample weights.
Returns
-------
score : float
R^2 of self.predict(X) wrt. y.
"""
y_pred = self.predict(K)
if sample_weight is None:
return np.mean(
[accuracy_score(y[j], y_pred[j]) for j in range(len(K))])
else:
return np.mean([
accuracy_score(y[j], y_pred[j], sample_weight=sample_weight[j])
for j in range(len(K))
])
def predict_proba(self, X):
"""Probability estimates.
The returned estimates for all classes are ordered by the
label of classes.
For a multi_class problem, if multi_class is set to be "multinomial"
the softmax function is used to find the predicted probability of
each class.
Else use a one-vs-rest approach, i.e calculate the probability
of each class assuming it to be positive using the logistic function.
and normalize these values across all the classes.
Parameters
----------
X : array-like, shape = [n_samples, n_features]
Returns
-------
T : array-like, shape = [n_samples, n_classes]
Returns the probability of the sample for each class in the model,
where classes are ordered as they are in ``self.classes_``.
"""
check_is_fitted(self, ["alpha_", "coef_"])
# return [LinearClassifierMixin._predict_proba_lr(
# self, np.tensordot(k, a, axes=1)) for a, k in zip(
# self.alpha_, X)]
return [
self.lr_p2[i].predict_proba(
np.tensordot(self.coef_.ravel(), X[i], axes=1))
for i in range(len(X))
]
def predict_log_proba(self, X):
"""Log of probability estimates.
The returned estimates for all classes are ordered by the
label of classes.
Parameters
----------
X : array-like, shape = [n_samples, n_features]
Returns
-------
T : array-like, shape = [n_samples, n_classes]
Returns the log-probability of the sample for each class in the
model, where classes are ordered as they are in ``self.classes_``.
"""
proba = self.predict_proba(X)
return [np.log(p) for p in proba]
3.0
>>> round_of_rating(4.1)
4.0"""
return np.round(number * 2) / 2
min_val = -40
max_val = 40
y_train = round_of_rating(saturate(y_train, min_val, max_val))
r_int = 0.5
slist = np.arange(min_val, max_val + r_int,
r_int) * 2 #multiply by 2 to allow labelbinarizer to work
lb = LabelBinarizer()
lb.fit(slist)
ylabels = lb.transform(y_train * 2)
# In[17]:
print(x_train.shape)
print(xfcss_train.shape)
print(ylabels.shape)
# In[18]:
nsamps = x_train.shape[0]
n80p = int(np.floor(nsamps * 0.8))
rannums = np.array(random.sample(range(1, nsamps, 1), n80p))
s_nfiles = np.arange(nsamps)
def fit(self, X, y=None):
if self.op == 'month' and hasattr(X, 'date'):
month = X.date.map(lambda x: x.month)
self.lb = LabelBinarizer()
self.lb.fit(month)
return self
class Model(ModelBase):
def __init__(self):
super(Model,self).__init__("LR")
self.vectorizer=CountVectorizer(lowercase=False,binary=True, analyzer=analyzer_)
self.binarizer=LabelBinarizer()
self.algo = LogisticRegression(C=1,fit_intercept=False,class_weight='balanced')
self.label_dict={}
self.label_index=0
self.label_map={}
def process_text(self, text):
words = text.lower().split()
stopwords = set('for a an the of and to in'.split())
words=[word for word in words if word not in stopwords]
return words
def hot_encode(self,word):
index=self.label_index
if(word[0] not in self.label_dict):
self.label_dict[word[0]]=self.label_index
self.label_map[self.label_index]=word[0]
index=self.label_index
self.label_index+=1
else:
index=self.label_dict[word[0]]
return index
#@ModelBase.train
def train(self,docs):
# Create sentences from documents
#docs=json.loads(data)
inputs=[]
labels=[]
for doc in docs:
inputs.append(self.process_text(doc['rawText']))
labels.append(self.hot_encode(doc['tags']))
self.binarizer.fit_transform(labels)
inputs = self.vectorizer.fit_transform(inputs)
print(inputs)
print(labels)
self.vectorizer.stop_words_=None
self.algo.fit(inputs, labels)
print('Training done')
return {'status': 2, 'reason': '', 'numRecords': len(inputs)}
def validate(self,docs):
#docs = json.loads(data)
inputs = []
labels = []
for doc in docs:
inputs.append(self.process_text(doc['rawText']))
labels.append(self.hot_encode(doc['tags']))
inputs = self.vectorizer.transform(inputs)
predictions = self.algo.predict(inputs)
fpr, tpr, threshold = roc_curve(labels, predictions)
roc = {"fpr": fpr.tolist(), "tpr": tpr.tolist(), "threshold": threshold.tolist()}
report = {
"classification_report": classification_report(labels, predictions),
"confusion_matrix": confusion_matrix(labels, predictions).tolist(),
"roc": roc,
"roc_auc": auc(fpr, tpr)
}
print (report)
return {'status': 2, 'reason':'', 'report': report, 'numRecords': len(inputs)}
def predict(self,docs):
#docs = json.loads(data)
inputs = [self.process_text(element['rawText']) for element in docs]
inputs = self.vectorizer.transform(inputs)
predictions = self.algo.predict(inputs)
results=[]
index=0
for i in predictions.tolist():
result={}
result["predicted_tags"]=self.label_map[i]
results.append(result)
index+=1
print (results)
return results
def persist(self, path):
package={
'vectorizer': self.vectorizer,
'binarizer': self.binarizer,
'algo': self.algo,
'label_map': self.label_map,
'label_dict': self.label_dict
}
with open(path, 'wb') as file:
pickle.dump(package,file, -1)
def reload(self, path):
with open(path, 'rb') as file:
package = pickle.load(file)
for key,value in package.items():
setattr(self,key,value)
def getFMFTRL():
#os.chdir('/Users/dhanley2/Documents/mercari/data')
os.chdir('/home/darragh/mercari/data')
train = pd.read_csv('../data/train.tsv', sep='\t', encoding='utf-8')
test = pd.read_csv('../data/test.tsv', sep='\t', encoding='utf-8')
glove_file = '../feat/glove.6B.50d.txt'
threads = 8
save_dir = '../feat'
print('[{}] Finished to load data'.format(time.time() - start_time))
print('Train shape: ', train.shape)
print('Test shape: ', test.shape)
nrow_test = train.shape[0] # -dftt.shape[0]
dftt = train[(train.price < 1.0)]
train = train.drop(train[(train.price < 1.0)].index)
del dftt['price']
nrow_train = train.shape[0]
# print(nrow_train, nrow_test)
y = np.log1p(train["price"])
merge = pd.concat([train, dftt, test])
merge['target'] = np.log1p(merge["price"])
submission = test[['test_id']]
ix = (merge['brand_name'] == merge['brand_name']) & (
~merge['brand_name'].str.lower().fillna('ZZZZZZ').isin(
merge['name'].str.lower()))
merge['name'][ix] = merge['brand_name'][ix] + ' ' + merge['name'][ix]
#EXTRACT DEVELOPTMENT TEST
trnidx, validx = train_test_split(range(train.shape[0]),
random_state=233,
train_size=0.90)
del train
del test
gc.collect()
merge['general_cat'], merge['subcat_1'], merge['subcat_2'] = \
zip(*merge['category_name'].apply(lambda x: split_cat(x)))
#merge.drop('category_name', axis=1, inplace=True)
print('[{}] Split categories completed.'.format(time.time() - start_time))
handle_missing_inplace(merge)
print('[{}] Handle missing completed.'.format(time.time() - start_time))
cutting(merge)
print('[{}] Cut completed.'.format(time.time() - start_time))
to_categorical(merge)
print('[{}] Convert categorical completed'.format(time.time() -
start_time))
'''
Crossed columns
'''
# my understanding on how to replicate what layers.crossed_column does. One
# can read here: https://www.tensorflow.org/tutorials/linear.
def cross_columns(x_cols):
"""simple helper to build the crossed columns in a pandas dataframe
"""
crossed_columns = dict()
colnames = ['_'.join(x_c) for x_c in x_cols]
for cname, x_c in zip(colnames, x_cols):
crossed_columns[cname] = x_c
return crossed_columns
merge['item_condition_id_str'] = merge['item_condition_id'].astype(str)
merge['shipping_str'] = merge['shipping'].astype(str)
x_cols = (
['brand_name', 'item_condition_id_str'],
['brand_name', 'subcat_1'],
['brand_name', 'subcat_2'],
['brand_name', 'general_cat'],
#['brand_name', 'subcat_1', 'item_condition_id_str'],
#['brand_name', 'subcat_2', 'item_condition_id_str'],
#['brand_name', 'general_cat', 'item_condition_id_str'],
['brand_name', 'shipping_str'],
['shipping_str', 'item_condition_id_str'],
['shipping_str', 'subcat_2'],
['item_condition_id_str', 'subcat_2'])
crossed_columns_d = cross_columns(x_cols)
categorical_columns = list(merge.select_dtypes(include=['object']).columns)
D = 2**30
for k, v in crossed_columns_d.items():
print('Crossed column ', k)
outls_ = []
indicator = 0
for col in v:
outls_.append((np.array(merge[col].apply(hash))) % D + indicator)
indicator += 10**6
merge[k] = sum(outls_).tolist()
'''
Count crossed cols
'''
cross_nm = [k for k in crossed_columns_d.keys()]
lb = LabelBinarizer(sparse_output=True)
x_col = lb.fit_transform(merge[cross_nm[0]])
for i in range(1, len(cross_nm)):
x_col = hstack((x_col, lb.fit_transform(merge[cross_nm[i]])))
del (lb)
gc.collect()
cpuStats()
'''
Hash name
'''
wb = wordbatch.WordBatch(normalize_text,
extractor=(WordBag, {
"hash_ngrams": 2,
"hash_ngrams_weights": [1.5, 1.0],
"hash_size": 2**29,
"norm": None,
"tf": 'binary',
"idf": None,
}),
procs=8)
wb.dictionary_freeze = True
X_name = wb.fit_transform(merge['name'])
del (wb)
X_name = X_name[:,
np.
array(np.clip(X_name.getnnz(axis=0) -
1, 0, 1), dtype=bool)]
print('[{}] Vectorize `name` completed.'.format(time.time() - start_time))
'''
Hash category
'''
wb = wordbatch.WordBatch(normalize_text,
extractor=(WordBag, {
"hash_ngrams": 2,
"hash_ngrams_weights": [1.0, 1.0],
"hash_size": 2**20,
"norm": None,
"tf": 'binary',
"idf": None,
}),
procs=8)
wb.dictionary_freeze = True
cat = merge["category_name"].str.replace('/', ' ')
X_cat = wb.fit_transform(cat)
del (wb)
X_cat = X_cat[:,
np.array(np.clip(X_cat.getnnz(axis=0) -
1, 0, 1), dtype=bool)]
print('[{}] Vectorize `category` completed.'.format(time.time() -
start_time))
'''
Count category
'''
wb = CountVectorizer()
X_category1 = wb.fit_transform(merge['general_cat'])
X_category2 = wb.fit_transform(merge['subcat_1'])
X_category3 = wb.fit_transform(merge['subcat_2'])
print('[{}] Count vectorize `categories` completed.'.format(time.time() -
start_time))
# wb= wordbatch.WordBatch(normalize_text, extractor=(WordBag, {"hash_ngrams": 3, "hash_ngrams_weights": [1.0, 1.0, 0.5],
wb = wordbatch.WordBatch(normalize_text,
extractor=(WordBag, {
"hash_ngrams": 2,
"hash_ngrams_weights": [1.0, 1.0],
"hash_size": 2**28,
"norm": "l2",
"tf": 1.0,
"idf": None
}),
procs=8)
wb.dictionary_freeze = True
X_description = wb.fit_transform(merge['item_description'])
del (wb)
X_description = X_description[:,
np.array(np.clip(
X_description.getnnz(axis=0) - 1, 0, 1),
dtype=bool)]
print('[{}] Vectorize `item_description` completed.'.format(time.time() -
start_time))
lb = LabelBinarizer(sparse_output=True)
X_brand = lb.fit_transform(merge['brand_name'])
print('[{}] Label binarize `brand_name` completed.'.format(time.time() -
start_time))
X_dummies = csr_matrix(
pd.get_dummies(merge[['item_condition_id', 'shipping']],
sparse=True).values)
print('[{}] Get dummies on `item_condition_id` and `shipping` completed.'.
format(time.time() - start_time))
'''
print(X_dummies.shape, X_description.shape, X_brand.shape, X_category1.shape, X_category2.shape, X_category3.shape,
X_name.shape, X_cat.shape, x_col.shape, X_orig.shape)
sparse_merge = hstack((X_dummies, X_description, X_brand, X_category1, X_category2, X_category3, X_name, X_cat,
x_col, X_orig)).tocsr()
'''
print(X_dummies.shape, X_description.shape, X_brand.shape,
X_category1.shape, X_category2.shape, X_category3.shape,
X_name.shape, X_cat.shape, x_col.shape)
sparse_merge = hstack(
(X_dummies, X_description, X_brand, X_category1, X_category2,
X_category3, X_name, X_cat, x_col)).tocsr()
print('[{}] Create sparse merge completed'.format(time.time() -
start_time))
# Remove features with document frequency <=1
print(sparse_merge.shape)
mask = np.array(np.clip(sparse_merge.getnnz(axis=0) - 1, 0, 1), dtype=bool)
sparse_merge = sparse_merge[:, mask]
X = sparse_merge[:nrow_train]
X_test = sparse_merge[nrow_test:]
print(sparse_merge.shape)
gc.collect()
if develop:
#train_X1, valid_X1, train_y1, valid_y1 = train_test_split(X, y, train_size=0.90, random_state=233)
train_X, valid_X, train_y, valid_y = X[trnidx], X[validx], y.values[
trnidx], y.values[validx]
model = FM_FTRL(alpha=0.01,
beta=0.01,
L1=0.00001,
L2=0.1,
D=sparse_merge.shape[1],
alpha_fm=0.01,
L2_fm=0.0,
init_fm=0.01,
D_fm=200,
e_noise=0.0001,
iters=1,
inv_link="identity",
threads=threads) #iters=15
baseline = 1.
for i in range(15):
model.fit(train_X, train_y, verbose=1)
predsfm = model.predict(X=valid_X)
score_ = rmsle(np.expm1(valid_y), np.expm1(predsfm))
print("FM_FTRL dev RMSLE:", score_)
if score_ < baseline - 0.0002:
baseline = score_
else:
break
print('[{}] Train ridge v2 completed'.format(time.time() - start_time))
if develop:
predsfm = model.predict(X=valid_X)
print("FM_FTRL dev RMSLE:", rmsle(np.expm1(valid_y),
np.expm1(predsfm)))
# 0.44532
# Full data 0.424681
predsFM = model.predict(X_test)
print('[{}] Predict FM_FTRL completed'.format(time.time() - start_time))
from wordbatch.models import nn_relu_h1, nn_relu_h2
modelnn = nn_relu_h1.NN_ReLU_H1(alpha=0.05, L2=0.00001, D_nn=60, D=sparse_merge.shape[1], \
iters=1, inv_link="identity", threads=threads)
baseline = 1.
print('[{}] Epoch time '.format(time.time() - start_time))
for i in range(3):
modelnn.fit(train_X, train_y, verbose=1)
predsnn = modelnn.predict(X=valid_X)
score_ = rmsle(np.expm1(valid_y), np.expm1(predsnn))
print("FM_FTRL dev RMSLE:", score_)
print('[{}] Epoch time '.format(time.time() - start_time))
if score_ < baseline - 0.0002:
baseline = score_
else:
break
pd.Series((np.expm1(predsnn) - np.expm1(predsfm))).hist()
print(
"FM_FTRL dev RMSLE:",
rmsle(np.expm1(valid_y),
0.1 * (np.expm1(predsnn)) + 0.9 * (np.expm1(predsfm))))
tpoint2 = time.time()
print("Time for Training: {}".format(hms_string(tpoint2 - tpoint1)))
return merge, trnidx, validx, nrow_train, nrow_test, glove_file, predsFM, predsfm
# 숫자 범주화
from sklearn.preprocessing import LabelEncoder
le = LabelEncoder()
le.fit([1, 2, 2, 6])
le.classes_ # scikit 에서 결과변수인 경우, '_'를 붙이는 관행이 있다.
le.transform([1, 1, 2, 6]) # labeling 하라
le.fit(["서울", "서울", "대전", "부산"])
le.classes_
le.transform(["서울", "서울", "부산"])
# 0과 1로 레이블링하라.
from sklearn.preprocessing import LabelBinarizer
lb = LabelBinarizer()
lb.fit([1, 2, 6, 1, 2]) # 1, 2, 6 의 세가지 경우가 있다.
lb.classes_
lb.transform([1, 6]) #1을 표현하는 것과, 6을 표현하는 것을 보여줌.
# dictionary feature 정보를 matrix로 표현.
from sklearn.feature_extraction import DictVectorizer
v = DictVectorizer(sparse=False)
D = [{
'foo': 1,
'bar': 2
}, {
'foo': 3,
'baz': 1
}] #한 문서에 foo가 1번, bar가 2번 나오고, 다른 문서에는 foo 3번, baz 1번.
def plot_mc_roc(y_test, y_score, interpreter=None):
'''
plotting function that generates roc curves for data given to it.
:param y_test: is the testing data used
:param y_score: is the score when the testing data was called
:param interpreter: is what was used to preprocess
:return a roc plot
'''
lw = 2
n_classes = len(np.unique(y_test))
classes = pd.unique(y_test)
label_binarizer = LabelBinarizer()
label_binarizer.fit(np.concatenate((y_test, y_score)))
if n_classes != 2:
y_test = label_binarizer.transform(y_test)
y_score = label_binarizer.transform(y_score)
else:
n_classes = 1
y_test = y_test.reshape(-1, 1)
y_score = y_score.reshape(-1, 1)
# Compute ROC curve and ROC area for each class
fpr = dict()
tpr = dict()
roc_auc = dict()
for i in range(n_classes):
fpr[i], tpr[i], _ = roc_curve(y_test[:, i], y_score[:, i])
roc_auc[i] = auc(fpr[i], tpr[i])
# Compute micro-average ROC curve and ROC area
fpr["micro"], tpr["micro"], _ = roc_curve(y_test.ravel(), y_score.ravel())
roc_auc["micro"] = auc(fpr["micro"], tpr["micro"])
# First aggregate all false positive rates
all_fpr = np.unique(np.concatenate([fpr[i] for i in range(n_classes)]))
# Then interpolate all ROC curves at this points
mean_tpr = np.zeros_like(all_fpr)
for i in range(n_classes):
mean_tpr += interp(all_fpr, fpr[i], tpr[i])
# Finally average it and compute AUC
mean_tpr /= n_classes
fpr["macro"] = all_fpr
tpr["macro"] = mean_tpr
roc_auc["macro"] = sklearn.metrics.auc(fpr["macro"], tpr["macro"])
# Plot all ROC curves
img = plt.figure()
plt.plot(fpr["micro"], tpr["micro"],
label='micro-average ROC curve (area = {0:0.2f})'
''.format(roc_auc["micro"]),
color='deeppink', linestyle=':', linewidth=4)
plt.plot(fpr["macro"], tpr["macro"],
label='macro-average ROC curve (area = {0:0.2f})'
''.format(roc_auc["macro"]),
color='navy', linestyle=':', linewidth=4)
for i in range(n_classes):
plt.plot(
fpr[i],
tpr[i],
lw=lw,
label='ROC curve of class {0} (area = {1:0.2f})'
''.format(
interpreter.inverse_transform(
[[label_binarizer.classes_[i]]])[0],
roc_auc[i]))
plt.plot([0, 1], [0, 1], 'k--', lw=lw)
plt.xlim([0.0, 1.0])
plt.ylim([0.0, 1.05])
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
plt.title('ROC Curves')
plt.legend(loc="lower right")
return img
from sklearn.preprocessing import StandardScaler
import random
import numpy as np
n_batch = 100
n_step = 8
n_input = 1
n_output = 10
n_cell = 100
lr = 0.006
n_train = 8000
bear = Bear()
data = scio.loadmat('wavedata.mat')['wavedata']
target = bear.target
lb = LabelBinarizer()
target = lb.fit_transform(target)
# print(target, ...)
# print(lb.inverse_transform(target))
X_train, X_test, y_train, y_test = train_test_split(data,
target,
test_size=0.333)
# scaler = StandardScaler()
# X_train = scaler.fit_transform(X_train)
# X_test = scaler.transform(X_test)
total_train_batch, total_test_batch = X_train.shape[0], X_test.shape[0]
print(total_train_batch, ...)
print(X_train.shape, y_test.shape, ...)
lstm = LSTM(n_batch, n_step, n_input, n_output, n_cell)
with tf.Session() as sess:
aap = AspectAwarePreprocessor(64, 64)
iap = ImageToArrayPreprocessor()
#load the dataset from disk and then scale the raw pixel intensities
#to the range [0,1]
sdl = SimpleDatasetLoader(preprocessors=[aap, iap])
(data, labels) = sdl.load(imagePaths, verbose=-1)
data = data.astype("float") / 255.0
(trainX, testX, trainY, testY) = train_test_split(data,
labels,
test_size=0.25,
random_state=42)
# convert the labels from integers to vectors
trainY = LabelBinarizer().fit_transform(trainY)
testY = LabelBinarizer().fit_transform(testY)
# initialize the optimizer and model
print("[INFO] compiling model...")
opt = SGD(lr=0.01, decay=0.01 / 20, momentum=0.9, nesterov=True)
model = Network.build(width=64, height=64, depth=3, classes=len(classNames))
model.compile(loss="categorical_crossentropy",
optimizer=opt,
metrics=["accuracy"])
# train the network
print("[INFO] training network...")
H = model.fit(trainX,
trainY,
validation_data=(testX, testY),
df_test = create_df(os.path.join(datapath, test_fname),
img_path,
partial_dataset=part_dat,
seed=123)
preds = np.load('10-02-2020_cont_colab.npy')
predsfl = np.load('22-01-2020_cont_colab.npy')
yhat = np.argmax(preds, axis=1) + 1
yhatfl = np.argmax(predsfl, axis=1) + 1
from sklearn.preprocessing import LabelBinarizer
y = df_test['label'].to_numpy()
lb = LabelBinarizer().fit(range(1, 40))
yhot = lb.transform(y)
#%%
from sklearn.metrics import roc_curve, auc, roc_auc_score
from scipy import interp
fpr = dict()
tpr = dict()
roc_auc = dict()
for i in range(39):
fpr[i], tpr[i], _ = roc_curve(yhot[:, i], preds[:, i])
if np.isnan(tpr[i]).any() or np.isnan(fpr[i]).any():
fpr[i] = tpr[i] = np.zeros(39)
roc_auc[i] = auc(fpr[i], tpr[i])
def load_data():
try:
# Reload the data from saved pickle file
with open(pickle_file, 'rb') as f:
pickle_data = pickle.load(f)
train_features_ = pickle_data['train_dataset']
train_labels_ = pickle_data['train_labels']
valid_features_ = pickle_data['valid_dataset']
valid_labels_ = pickle_data['valid_labels']
test_features_ = pickle_data['test_dataset']
test_labels_ = pickle_data['test_labels']
del pickle_data # Free up memory
return train_features_, train_labels_, valid_features_, valid_labels_, test_features_, test_labels_
except IOError:
print("failed to reload data from %s, load it from begining" %
pickle_file)
# Download the training and test dataset.
download_if_necessary(
'https://s3.amazonaws.com/udacity-sdc/notMNIST_train.zip',
'notMNIST_train.zip', 'c8673b3f28f489e9cdf3a3d74e2ac8fa')
download_if_necessary(
'https://s3.amazonaws.com/udacity-sdc/notMNIST_test.zip',
'notMNIST_test.zip', '5d3c7e653e63471c88df796156a9dfa9')
# Make sure the files aren't corrupted
assert hashlib.md5(open('notMNIST_train.zip', 'rb').read()).hexdigest() == 'c8673b3f28f489e9cdf3a3d74e2ac8fa', \
'notMNIST_train.zip file is corrupted. Remove the file and try again.'
assert hashlib.md5(open('notMNIST_test.zip', 'rb').read()).hexdigest() == '5d3c7e653e63471c88df796156a9dfa9', \
'notMNIST_test.zip file is corrupted. Remove the file and try again.'
# Get the features and labels from the zip files
train_features_, train_labels_ = uncompress_features_labels(
'notMNIST_train.zip')
test_features_, test_labels_ = uncompress_features_labels(
'notMNIST_test.zip')
# Limit the amount of data to work with a docker container
docker_size_limit = 150000
train_features_, train_labels_ = resample(train_features_,
train_labels_,
n_samples=docker_size_limit)
# normalize the data
train_features_ = normalize_grayscale(train_features_)
test_features_ = normalize_grayscale(test_features_)
# Turn labels into numbers and apply One-Hot Encoding
encoder = LabelBinarizer()
encoder.fit(train_labels_)
# one-hor encode, and change to float32,
# so it can be multiplied against the features in TensorFlow, which are float32
train_labels_ = encoder.transform(train_labels_).astype(np.float32)
test_labels_ = encoder.transform(test_labels_).astype(np.float32)
# Get randomized datasets for training and validation
train_features_, valid_features_, train_labels_, valid_labels_ = train_test_split(
train_features_, train_labels_, test_size=0.05, random_state=832289)
# Save the data for easy access
if not os.path.isfile(pickle_file):
print('Saving data to pickle file...')
try:
with open('notMNIST.pickle', 'wb') as pfile:
pickle.dump(
{
'train_dataset': train_features_,
'train_labels': train_labels_,
'valid_dataset': valid_features_,
'valid_labels': valid_labels_,
'test_dataset': test_features_,
'test_labels': test_labels_,
}, pfile, pickle.HIGHEST_PROTOCOL)
print('Data cached in pickle file.')
except Exception as e:
print('Unable to save data to', pickle_file, ':', e)
raise
return train_features_, train_labels_, valid_features_, valid_labels_, test_features_, test_labels_
import tensorflow as tf
from sklearn.datasets import load_digits
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelBinarizer
# load data
digits = load_digits()
x_set = digits.data
y_set = LabelBinarizer().fit_transform(digits.target)
x_train, x_test, y_train, y_test = train_test_split(x_set,
y_set,
test_size=0.3)
def add_layer(inputs, in_size, out_size, activition_function=None):
Weights = tf.Variable(tf.random_normal([in_size, out_size]))
biases = tf.Variable(tf.zeros([1, out_size]) + 0.1) # 不推荐为0
Wx_plus_b = tf.matmul(inputs, Weights) + biases
Wx_plus_b = tf.nn.dropout(Wx_plus_b, keep_probability)
if activition_function is None:
outputs = Wx_plus_b
else:
outputs = activition_function(Wx_plus_b)
tf.summary.histogram('/outputs', outputs)
return outputs
keep_probability = tf.placeholder(tf.float32) # 保存率
X = tf.placeholder(tf.float32, shape=[None, 64])
Y = tf.placeholder(tf.float32, shape=[None, 10])
def eval_models(model_paths, data_path, save_path=None):
'''
Evaluates performance of a model in terms of loss, accuracy, confusion
matrix, and mean per-class recall
Parameters:
model_paths(dict) - dictionary with model names as keys and paths pointing
to the .h5 files of the trained models as values
data_path(string) - path to the image directory of the target dataset
json_path(string) - optional file path to save output to
Returns:
Dictionary of dictionaries each containing loss, accuracy, confusion matrix,
and mean per-class recall for a given model
'''
from keras.models import load_model
from keras.backend import clear_session
from keras.preprocessing.image import ImageDataGenerator
from sklearn.metrics import confusion_matrix, f1_score, roc_auc_score
from sklearn.preprocessing import LabelBinarizer
import gc
import json
# build generator to feed the model
print('Building image generator...')
generator = ImageDataGenerator().flow_from_directory(data_path,
target_size=(224,
224),
batch_size=8,
shuffle=False)
y_true = generator.classes
# evaluate all models
model_results = dict()
for name, path in model_paths.items():
# load model
print('Loading {}'.format(path))
model = load_model(path)
# run basic evaluation
print('Evaluating {}'.format(path))
metrics = dict()
metrics['loss'], metrics['acc'] = model.evaluate_generator(generator)
# predict labels
y_prob = model.predict_generator(generator)
y_pred = y_prob.argmax(axis=1)
# calculate confusion matrix
cm = confusion_matrix(y_true, y_pred)
metrics['cm'] = cm.tolist()
# mean per-class recall
pcr = cm.diagonal() / cm.sum(axis=1)
metrics['pcr'] = list(pcr)
metrics['mpcr'] = pcr.mean()
# F1 score
metrics['class_f1s'] = list(f1_score(y_true, y_pred, average=None))
metrics['macro_f1'] = f1_score(y_true, y_pred, average='macro')
metrics['micro_f1'] = f1_score(y_true, y_pred, average='micro')
# AUC score
y_binary = LabelBinarizer().fit_transform(y_true)
metrics['class_aucs'] = list(
roc_auc_score(y_binary, y_prob, average=None))
metrics['macro_auc'] = roc_auc_score(y_binary, y_prob, average='macro')
metrics['micro_auc'] = roc_auc_score(y_binary, y_prob, average='micro')
model_results[name] = metrics
# remove clutter from memory
del model
clear_session()
gc.collect()
if save_path:
print('Saving evaluation to {}'.format(save_path))
with open(save_path, 'w') as f:
json.dump(model_results, f)
print('Evaluation complete.\n')
gc.collect()
return model_results
class MatheusAlvesMLP(BaseEstimator, ClassifierMixin): # or RegressonMixin?
def __init__(self, params=None):
if params is None:
self.ctor({})
else:
self.ctor(params)
def ctor(self, params):
self.alpha = params.get("alpha", 0.00001) # L2 regularization
self.max_iter = params.get(
"max_iter", 500) # max iteration to the optimization algorithm
self.hidden_layers_size = params.get("hidden_layers_size",
(100, 200, 300))
self.shuffle = params.get("shuffle",
False) # shuflle samples in interactions?
self.random_state = params.get(
"random_state", None) # state or seed for generating random number
self.tol = params.get("tol", 1e-5) # Loss tolerance for optimization
self.layers_coef = None
self.layers_intercept = None
self.cost = None
self.n_iter = 0
self.classes = None
self.label_binarizer_ = LabelBinarizer()
def _unstack(self, stacked_parameters):
for i in range(self.n_layers_ - 1):
start, end, shape = self._coef_indptr[i]
self.layers_coef[i] = np.reshape(stacked_parameters[start:end],
shape)
start, end = self._intercept_indptr[i]
self.layers_intercept[i] = stacked_parameters[start:end]
def _forward_pass(self, activations, with_output_activation=True):
for i in range(self.n_layers_ - 1):
activations[i + 1] = safe_sparse_dot(activations[i],
self.layers_coef[i])
activations[i + 1] += self.layers_intercept[i]
# For the hidden layers
if i + 1 != self.n_layers_ - 1:
activations[i + 1] = rectified_linear_unit(activations[i + 1])
# For the last layer
if with_output_activation:
activations[i + 1] = rectified_linear_unit(activations[i + 1])
return activations
def _compute_cost_grad(self, layer, n_samples, activations, deltas,
coef_grads, intercept_grads):
coef_grads[layer] = safe_sparse_dot(activations[layer].T,
deltas[layer])
coef_grads[layer] += (self.alpha * self.layers_coef[layer])
coef_grads[layer] /= n_samples
intercept_grads[layer] = np.mean(deltas[layer], 0)
return coef_grads, intercept_grads
def _cost_grad_lbfgs(self, stacked_coef_inter, X, y, activations, deltas,
coef_grads, intercept_grads):
self._unstack(stacked_coef_inter)
cost, coef_grads, intercept_grads = self._backprop(
X, y, activations, deltas, coef_grads, intercept_grads)
self.n_iter += 1
grad = stack(coef_grads, intercept_grads)
return cost, grad
def _backprop(self, X, y, activations, deltas, coef_grads,
intercept_grads):
n_samples = X.shape[0]
# Forward propagate
activations = self._forward_pass(activations)
# Get cost using log loss function
cost = log_loss(y, activations[-1])
# Add regularization term to the cost
values = np.sum(np.array([np.sum(s**2) for s in self.layers_coef]))
cost += (0.5 * self.alpha) * values / n_samples
# Backward propagate
last = self.n_layers_ - 2
diff = y - activations[-1]
deltas[last] = -diff
# Compute gradient for the last layer
coef_grads, intercept_grads = self._compute_cost_grad(
last, n_samples, activations, deltas, coef_grads, intercept_grads)
# Iterate over the hidden layers
for i in range(self.n_layers_ - 2, 0, -1):
deltas[i - 1] = safe_sparse_dot(deltas[i], self.layers_coef[i].T)
deltas[i - 1] *= rectified_linear_unit_derivative(activations[i])
coef_grads, intercept_grads = self._compute_cost_grad(
i - 1, n_samples, activations, deltas, coef_grads,
intercept_grads)
return cost, coef_grads, intercept_grads
def fit(self, X, y):
hidden_layers_size = list(self.hidden_layers_size)
n_samples, n_features = X.shape
self.label_binarizer_.fit(y)
if self.classes is None:
self.classes = self.label_binarizer_.classes_
else:
classes = self.label_binarizer_.classes_
y = self.label_binarizer_.transform(y)
self.n_outputs = y.shape[1]
layer_units = ([n_features] + hidden_layers_size + [self.n_outputs])
# If it is the first time training the model
if self.layers_coef is None:
# Initialize parameters
self.n_outputs = y.shape[1]
# Compute the number of layers
self.n_layers_ = len(layer_units)
# Initialize coefficient and intercept layers
self.layers_coef = []
self.layers_intercept = []
for i in range(self.n_layers_ - 1):
rng = check_random_state(self.random_state)
n_fan_in = layer_units[i]
n_fan_out = layer_units[i + 1]
# Use the Gorot initialization method
weight_init_bound = np.sqrt(6. / (n_fan_in + n_fan_out))
self.layers_coef.append(
rng.uniform(-weight_init_bound, weight_init_bound,
(n_fan_in, n_fan_out)))
self.layers_intercept.append(
rng.uniform(-weight_init_bound, weight_init_bound,
n_fan_out))
if self.shuffle:
X, y = shuffle(X, y, random_state=self.random_state)
# Initialize lists
activations = [X]
activations.extend(
np.empty((n_samples, n_fan_out)) for n_fan_out in layer_units[1:])
deltas = [np.empty_like(a_layer) for a_layer in activations]
coef_grads = [
np.empty((n_fan_in, n_fan_out))
for n_fan_in, n_fan_out in zip(layer_units[:-1], layer_units[1:])
]
intercept_grads = [
np.empty(n_fan_out) for n_fan_out in layer_units[1:]
]
# START LBFGS algorithm
# Store meta information for the parameters
self._coef_indptr = []
self._intercept_indptr = []
start = 0
# Save sizes and indices of coefficients for faster unstacking
for i in range(self.n_layers_ - 1):
n_fan_in, n_fan_out = layer_units[i], layer_units[i + 1]
end = start + (n_fan_in * n_fan_out)
self._coef_indptr.append((start, end, (n_fan_in, n_fan_out)))
start = end
# Save sizes and indices of intercepts for faster unstacking
for i in range(self.n_layers_ - 1):
end = start + layer_units[i + 1]
self._intercept_indptr.append((start, end))
start = end
# enable pretty output for l_bfgs_b
iprint = 1
# Run L-BFGS_B opitimization
stacked_coef_inter = stack(self.layers_coef, self.layers_intercept)
optimal_parameters, self.cost, d = fmin_l_bfgs_b(
x0=stacked_coef_inter,
func=self._cost_grad_lbfgs,
maxfun=self.max_iter,
iprint=iprint,
pgtol=self.tol,
args=(X, y, activations, deltas, coef_grads, intercept_grads))
self._unstack(optimal_parameters)
return self
def decision_function(self, X):
hidden_layers_size = list(self.hidden_layers_size)
layer_units = [X.shape[1]] + hidden_layers_size + [self.n_outputs]
# Initialize layers
activations = []
activations.append(X)
for i in range(self.n_layers_ - 1):
activations.append(np.empty((X.shape[0], layer_units[i + 1])))
# forward propagate
self._forward_pass(activations, with_output_activation=False)
y_pred = activations[-1]
if self.n_outputs == 1:
return y_pred.ravel()
else:
return y_pred
def predict(self, X):
y_scores = self.decision_function(X)
y_scores = rectified_linear_unit(y_scores)
return self.label_binarizer_.inverse_transform(y_scores)
def trainFMFTRL(moddict):
merge = pd.read_csv(trn_file, sep='\t', encoding='utf-8')
#test = pd.read_csv(tst_file, sep='\t', encoding='utf-8')
print('[{}] Finished to load data'.format(time.time() - start_time))
print('Train shape: ', merge.shape)
dftt = merge[(merge.price < 1.0)]
merge = merge.drop(merge[(merge.price < 1.0)].index)
del dftt['price']
nrow_train = merge.shape[0]
# print(nrow_train, nrow_test)
y = np.log1p(merge["price"])
merge = pd.concat([merge, dftt])
merge['target'] = np.log1p(merge["price"])
#EXTRACT DEVELOPTMENT TEST
trnidx, validx = train_test_split(range(merge[:nrow_train].shape[0]),
random_state=233,
train_size=0.90)
gc.collect()
cpuStats()
merge = prepFMFeatures(merge)
cpuStats()
merge.head()
'''
Count crossed cols
'''
cross_nm = [k for k in crossed_columns_d.keys()]
moddict['cross_cols'] = {}
for i in range(0, len(cross_nm)):
moddict['cross_cols'][cross_nm[i]] = LabelBinarizer(sparse_output=True)
moddict['cross_cols'][cross_nm[i]].fit(merge[cross_nm[i]])
if i == 0:
x_col = moddict['cross_cols'][cross_nm[i]].transform(
merge[cross_nm[i]])
else:
x_col = hstack(
(x_col, moddict['cross_cols'][cross_nm[i]].fit_transform(
merge[cross_nm[i]])))
del merge[cross_nm[i]]
gc.collect()
cpuStats()
'''
Hash name
'''
moddict['wb_name'] = wordbatch.WordBatch(normalize_text,
extractor=(WordBag, {
"hash_ngrams":
2,
"hash_ngrams_weights":
[1.5, 1.0],
"hash_size":
2**29,
"norm":
None,
"tf":
'binary',
"idf":
None,
'verbose':
1,
}),
procs=8)
moddict['wb_name'].dictionary_freeze = True
X_name = moddict['wb_name'].fit_transform(merge['name'])
moddict['wb_name_mask'] = np.where(
X_name[:nrow_train].getnnz(axis=0) > 0)[0]
X_name = X_name[:, moddict['wb_name_mask']]
print('[{}] Vectorize `name` completed.'.format(time.time() - start_time))
'''
Hash category #2
'''
moddict['wb_cat'] = wordbatch.WordBatch(normalize_text,
extractor=(WordBag, {
"hash_ngrams":
2,
"hash_ngrams_weights":
[1.0, 1.0],
"hash_size":
2**20,
"norm":
None,
"tf":
'binary',
"idf":
None,
}),
procs=4)
moddict['wb_cat'].dictionary_freeze = True
### This must be the full dataset
#cats = merge["category_name"].str.replace('/', ' ').unique()
moddict['wb_cat'].fit(categories)
X_cat_tmp = moddict['wb_cat'].transform(categories)
moddict['wb_cat_dict'] = dict([
(c, X_cat_tmp.getrow(row))
for (c, row) in zip(categories.tolist(), range(len(categories)))
])
X_cat = vstack(([
moddict['wb_cat_dict'][c]
for c in merge["category_name"].str.replace('/', ' ')
]))
#moddict['wb_cat_mask'] = np.array(np.clip(X_cat[:nrow_train].getnnz(axis=0) - 1, 0, 1), dtype=bool)
moddict['wb_cat_mask'] = np.where(X_cat[:nrow_train].getnnz(axis=0) > 0)[0]
X_cat = X_cat[:, moddict['wb_cat_mask']]
print('[{}] Vectorize `category` completed.'.format(time.time() -
start_time))
'''
Count category
'''
moddict['wb_cat_ctgc'] = CountVectorizer()
moddict['wb_cat_ctgc'].fit(merge['general_cat'])
X_category1 = moddict['wb_cat_ctgc'].transform(merge['general_cat'])
moddict['wb_cat_ctsc1'] = CountVectorizer()
moddict['wb_cat_ctsc1'].fit(merge['subcat_1'])
X_category2 = moddict['wb_cat_ctsc1'].transform(merge['subcat_1'])
moddict['wb_cat_ctsc2'] = CountVectorizer()
moddict['wb_cat_ctsc2'].fit(merge['subcat_2'])
X_category3 = moddict['wb_cat_ctsc2'].transform(merge['subcat_2'])
print('[{}] Count vectorize `categories` completed.'.format(time.time() -
start_time))
moddict['wb_dscr'] = wordbatch.WordBatch(normalize_text,
extractor=(WordBag, {
"hash_ngrams":
2,
"hash_ngrams_weights":
[1.0, 0.6],
"hash_size":
2**28,
"norm":
None,
"tf":
'binary',
"idf":
None
}),
procs=8)
moddict['wb_dscr'].dictionary_freeze = True
X_description = moddict['wb_dscr'].fit_transform(merge['name'] + ' ' +
merge['item_description'])
#moddict['wb_dscr_mask'] = np.array(np.clip(X_description[:nrow_train].getnnz(axis=0) - 1, 0, 1), dtype=bool)
moddict['wb_dscr_mask'] = np.where(
X_description[:nrow_train].getnnz(axis=0) > 1)[0]
X_description = X_description[:, moddict['wb_dscr_mask']]
print('[{}] Vectorize `item_description` completed.'.format(time.time() -
start_time))
moddict['wb_brandname'] = LabelBinarizer(sparse_output=True)
moddict['wb_brandname'].fit(merge['brand_name'][:nrow_train])
X_brand = moddict['wb_brandname'].transform(merge['brand_name'])
print('[{}] Label binarize `brand_name` completed.'.format(time.time() -
start_time))
moddict['wb_itemcond'] = LabelBinarizer(sparse_output=True)
moddict['wb_itemcond'].fit(merge['item_condition_id'][:nrow_train])
X_itemcond = moddict['wb_itemcond'].transform(merge['item_condition_id'])
print('[{}] Label binarize `item_condition_id` completed.'.format(
time.time() - start_time))
moddict['wb_shipping'] = LabelBinarizer(sparse_output=True)
moddict['wb_shipping'].fit(merge['shipping'][:nrow_train])
X_shipping = moddict['wb_shipping'].transform(merge['shipping'])
print('[{}] Label binarize `shipping` completed.'.format(time.time() -
start_time))
print(
X_itemcond.shape,
X_shipping.shape, #X_dummies.shape,
X_description.shape,
X_brand.shape,
X_category1.shape,
X_category2.shape,
X_category3.shape,
X_name.shape,
X_cat.shape,
x_col.shape)
sparse_merge = hstack((
X_itemcond,
X_shipping, #X_dummies,
X_description,
X_brand,
X_category1,
X_category2,
X_category3,
X_name,
X_cat,
x_col)).tocsr()
print('[{}] Create sparse merge completed'.format(time.time() -
start_time))
print(50 * '-')
cpuStats()
print(50 * '-')
# Remove features with document frequency <=1
print(sparse_merge.shape)
gc.collect()
sparse_merge, y = sparse_merge[:nrow_train], y[:nrow_train]
if develop:
train_X, valid_X, train_y, valid_y = sparse_merge[trnidx], \
sparse_merge[validx], \
y.values[trnidx], y.values[validx]
del sparse_merge
gc.collect()
print(50 * '*')
cpuStats()
print(50 * '*')
print(train_X.shape[1])
model = FM_FTRL(alpha=0.01,
beta=0.01,
L1=0.00001,
L2=0.1,
D=train_X.shape[1],
alpha_fm=0.01,
L2_fm=0.0,
init_fm=0.01,
D_fm=200,
e_noise=0.0001,
iters=1,
inv_link="identity",
threads=4) #iters=15
print(50 * '|')
cpuStats()
print(50 * '|')
baseline = 1.
for i in range(15):
print(50 * '-')
cpuStats()
print(50 * '-')
model.fit(train_X, train_y, verbose=1)
predsfm = model.predict(X=valid_X)
score_ = rmsle(np.expm1(valid_y), np.expm1(predsfm))
print("FM_FTRL dev RMSLE:", score_)
if score_ < baseline - 0.0004:
baseline = score_
else:
break
moddict['FMmodel'] = model
print('[{}] Train ridge v2 completed'.format(time.time() - start_time))
if develop:
predsfm = moddict['FMmodel'].predict(X=valid_X)
print("FM_FTRL dev RMSLE:", rmsle(np.expm1(valid_y),
np.expm1(predsfm)))
gc.collect()
return merge, moddict, trnidx, validx, nrow_train, predsfm
image_path.split(os.path.sep)[LABEL_PATH_INDEX]
for image_path in image_paths
]
class_names = [str(x) for x in np.unique(class_names)]
aap = AspectAwarePreprocessor(WIDTH, HEIGHT)
iap = ImageToArrayPreprocessor()
sdl = SimpleDatasetLoader([aap, iap])
data, labels = sdl.load(image_paths, verbose=500)
data = data.astype('float') / 255.0
train_X, test_X, train_y, test_y = train_test_split(data,
labels,
test_size=0.25)
label_binarizer = LabelBinarizer()
train_y = label_binarizer.fit_transform(train_y)
test_y = label_binarizer.transform(test_y)
aug = ImageDataGenerator(rotation_range=30,
width_shift_range=0.1,
height_shift_range=0.1,
shear_range=0.2,
zoom_range=0.2,
horizontal_flip=True,
fill_mode='nearest')
print('[INFO] compiling model')
model = MiniVGGNet.build(width=WIDTH,
height=HEIGHT,
depth=3,
train_posts = data['news'][:train_size]
train_tags = data['category'][:train_size]
train_files_names = data['filename'][:train_size]
test_posts = data['news'][train_size:]
test_tags = data['category'][train_size:]
test_files_names = data['filename'][train_size:]
# define Tokenizer with Vocab Size
tokenizer = Tokenizer(num_words=vocab_size)
tokenizer.fit_on_texts(train_posts)
x_train = tokenizer.texts_to_matrix(train_posts, mode='tfidf')
x_test = tokenizer.texts_to_matrix(test_posts, mode='tfidf')
encoder = LabelBinarizer()
encoder.fit(train_tags)
y_train = encoder.transform(train_tags)
y_test = encoder.transform(test_tags)
model = Sequential()
model.add(Dense(512, input_shape=(vocab_size,)))
model.add(Activation('relu'))
model.add(Dropout(0.3))
model.add(Dense(512))
model.add(Activation('relu'))
model.add(Dropout(0.3))
model.add(Dense(num_labels))
model.add(Activation('softmax'))
model.summary()
def test_gradient():
# Test gradient.
# This makes sure that the activation functions and their derivatives
# are correct. The numerical and analytical computation of the gradient
# should be close.
for n_labels in [2, 3]:
n_samples = 5
n_features = 10
random_state = np.random.RandomState(seed=42)
X = random_state.rand(n_samples, n_features)
y = 1 + np.mod(np.arange(n_samples) + 1, n_labels)
Y = LabelBinarizer().fit_transform(y)
for activation in ACTIVATION_TYPES:
mlp = MLPClassifier(activation=activation,
hidden_layer_sizes=10,
solver='lbfgs',
alpha=1e-5,
learning_rate_init=0.2,
max_iter=1,
random_state=1)
mlp.fit(X, y)
theta = np.hstack(
[l.ravel() for l in mlp.coefs_ + mlp.intercepts_])
layer_units = ([X.shape[1]] + [mlp.hidden_layer_sizes] +
[mlp.n_outputs_])
activations = []
deltas = []
coef_grads = []
intercept_grads = []
activations.append(X)
for i in range(mlp.n_layers_ - 1):
activations.append(np.empty((X.shape[0], layer_units[i + 1])))
deltas.append(np.empty((X.shape[0], layer_units[i + 1])))
fan_in = layer_units[i]
fan_out = layer_units[i + 1]
coef_grads.append(np.empty((fan_in, fan_out)))
intercept_grads.append(np.empty(fan_out))
# analytically compute the gradients
def loss_grad_fun(t):
return mlp._loss_grad_lbfgs(t, X, Y, activations, deltas,
coef_grads, intercept_grads)
[value, grad] = loss_grad_fun(theta)
numgrad = np.zeros(np.size(theta))
n = np.size(theta, 0)
E = np.eye(n)
epsilon = 1e-5
# numerically compute the gradients
for i in range(n):
dtheta = E[:, i] * epsilon
numgrad[i] = ((loss_grad_fun(theta + dtheta)[0] -
loss_grad_fun(theta - dtheta)[0]) /
(epsilon * 2.0))
assert_almost_equal(numgrad, grad)
def training_summary(H):
N = EPOCHS
plt.style.use("ggplot")
plt.figure()
plt.plot(np.arange(0, N), H.history["loss"], label="train_loss")
plt.plot(np.arange(0, N), H.history["val_loss"], label="val_loss")
plt.plot(np.arange(0, N), H.history["accuracy"], label="train_acc")
plt.plot(np.arange(0, N), H.history["val_accuracy"], label="val_acc")
plt.title("Training Loss and Accuracy")
plt.xlabel("Epoch #")
plt.ylabel("Loss/Accuracy")
plt.legend(loc="lower left")
plt.savefig(args["plot"])
if __name__ == '__main__':
INIT_LR = 1e-4
EPOCHS = 20
BS = 32
args = argument_parser()
lb = LabelBinarizer()
path = list(paths.list_images(args["dataset"]))
data, labels = get_images(path)
labels = one_hot_encoder(lb, labels)
aug = augmetation_image()
xtrain, xtest, ytrain, ytest = data_splitting(data, labels)
model, H = model_construction(aug, xtrain, xtest, ytrain, ytest)
training_summary(H)
