def randomize(self, data_matrix, amount=.5):
random.seed(self.random_state)
inclusion_threshold = random.uniform(amount, 1)
selectors = []
if random.random() > inclusion_threshold:
selectors.append(
SparseSelector(random_state=random.randint(1, 1e9)))
if random.random() > inclusion_threshold:
selectors.append(
MaxVolSelector(random_state=random.randint(1, 1e9)))
if random.random() > inclusion_threshold:
selectors.append(
QuickShiftSelector(random_state=random.randint(1, 1e9)))
if random.random() > inclusion_threshold:
selectors.append(
DensitySelector(random_state=random.randint(1, 1e9)))
if random.random() > inclusion_threshold:
selectors.append(
OnionSelector(random_state=random.randint(1, 1e9)))
if not selectors:
selectors.append(
DensitySelector(random_state=random.randint(1, 1e9)))
selectors.append(
SparseSelector(random_state=random.randint(1, 1e9)))
selector = CompositeSelector(selectors=selectors)
selector.randomize(data_matrix, amount=amount)
self.selectors = deepcopy(selector.selectors)
self.metric = 'rbf'
self.kwds = {'gamma': random.choice([10**x for x in range(-3, 3)])}
if random.random() > inclusion_threshold:
self.n_nearest_neighbors = random.randint(3, 20)
else:
self.n_nearest_neighbors = None
self.n_links = random.randint(1, 5)
self.random_state = self.random_state ^ random.randint(1, 1e9)
def randomize(self, data_matrix, amount=.5):
random.seed(self.random_state)
inclusion_threshold = random.uniform(amount, 1)
selectors = []
if random.random() > inclusion_threshold:
selectors.append(SparseSelector(random_state=random.randint(1, 1e9)))
if random.random() > inclusion_threshold:
selectors.append(MaxVolSelector(random_state=random.randint(1, 1e9)))
if random.random() > inclusion_threshold:
selectors.append(QuickShiftSelector(random_state=random.randint(1, 1e9)))
if random.random() > inclusion_threshold:
selectors.append(DensitySelector(random_state=random.randint(1, 1e9)))
if random.random() > inclusion_threshold:
selectors.append(OnionSelector(random_state=random.randint(1, 1e9)))
if not selectors:
selectors.append(DensitySelector(random_state=random.randint(1, 1e9)))
selectors.append(SparseSelector(random_state=random.randint(1, 1e9)))
selector = CompositeSelector(selectors=selectors)
selector.randomize(data_matrix, amount=amount)
self.selectors = deepcopy(selector.selectors)
self.metric = 'rbf'
self.kwds = {'gamma': random.choice([10 ** x for x in range(-3, 3)])}
if random.random() > inclusion_threshold:
self.n_nearest_neighbors = random.randint(3, 20)
else:
self.n_nearest_neighbors = None
self.n_nearest_neighbor_links = random.randint(1, 5)
self.random_state = self.random_state ^ random.randint(1, 1e9)
def feature_construction(data_matrix_original, selectors):
from eden.selector import CompositeSelector
selector = CompositeSelector(selectors)
from eden.selector import Projector
projector = Projector(selector, metric='cosine')
data_matrix = projector.fit_transform(data_matrix_original)
return data_matrix
class Projector(object):
"""Constructs features as the instance similarity to a set of instances as
defined by the selector.
Parameters
----------
selector : Selector
TODO.
scale : bool (default True)
If true then the data matrix returned is standardized to have 0 mean and unit variance
scaling_factor : float (default 0.8)
Multiplicative factor applied after normalization. This can be useful when data needs to be
post-processed by neural networks and one wishes to push data in a linear region.
random_state : int (deafault 1)
The seed used for the pseudo-random generator.
metric : string, or callable
The metric to use when calculating kernel between instances in a
feature array. If metric is a string, it must be one of the metrics
in pairwise.PAIRWISE_KERNEL_FUNCTIONS.
If metric is "precomputed", X is assumed to be a kernel matrix.
Alternatively, if metric is a callable function, it is called on each
pair of instances (rows) and the resulting value recorded. The callable
should take two arrays from X as input and return a value indicating
the distance between them.
**kwds : optional keyword parameters
Any further parameters are passed directly to the kernel function.
"""
def __init__(self,
selector=AllSelector(),
scale=True,
scaling_factor=0.8,
random_state=1,
metric='rbf',
**kwds):
self.selector = selector
self.scale = scale
self.scaling_factor = scaling_factor
self.scaler = StandardScaler()
self.metric = metric
self.kwds = kwds
self.random_state = random_state
def __repr__(self):
serial = []
serial.append('Projector:')
serial.append('metric: %s' % self.metric)
if self.kwds is None or len(self.kwds) == 0:
pass
else:
serial.append('params:')
serial.append(serialize_dict(self.kwds))
serial.append(str(self.selector))
return '\n'.join(serial)
def fit(self, data_matrix, target=None):
"""Fit the estimator on the samples.
Parameters
----------
data_matrix : array-like, shape = (n_samples, n_features)
Samples.
Returns
-------
self
"""
self.selected_instances = self.selector.fit_transform(data_matrix,
target=target)
if self.scale:
self.scale = False
self.scaler.fit(self.transform(data_matrix))
self.scale = True
return self
def fit_transform(self, data_matrix, target=None):
"""Fit the estimator on the samples and transforms features as the instance
similarity to a set of instances as defined by the selector.
Parameters
----------
data_matrix : array, shape = (n_samples, n_features)
Samples.
target : TODO
Returns
-------
data_matrix : array, shape = (n_samples, n_features_new)
Transformed array.
"""
self.fit(data_matrix, target)
return self.transform(data_matrix)
def transform(self, data_matrix):
"""Transforms features as the instance similarity to a set of instances as
defined by the selector.
Parameters
----------
data_matrix : array, shape = (n_samples, n_features)
Samples.
Returns
-------
data_matrix : array, shape = (n_samples, n_features_new)
Transformed array.
"""
if self.selected_instances is None:
raise Exception('Error: attempt to use transform on non fit model')
if self.selected_instances.shape[0] == 0:
raise Exception(
'Error: attempt to use transform using 0 selectors')
data_matrix_out = pairwise_kernels(data_matrix,
Y=self.selected_instances,
metric=self.metric,
**self.kwds)
if self.scale:
data_matrix_out = self.scaler.transform(
data_matrix_out) * self.scaling_factor
return data_matrix_out
def randomize(self, data_matrix, amount=.5):
random.seed(self.random_state)
inclusion_threshold = random.uniform(amount, 1)
selectors = []
if random.random() > inclusion_threshold:
selectors.append(
QuickShiftSelector(random_state=random.randint(1, 1e9)))
if random.random() > inclusion_threshold:
selectors.append(
DecisionSurfaceSelector(random_state=random.randint(1, 1e9)))
if random.random() > inclusion_threshold:
selectors.append(
SparseSelector(random_state=random.randint(1, 1e9)))
if random.random() > inclusion_threshold:
selectors.append(
MaxVolSelector(random_state=random.randint(1, 1e9)))
if random.random() > inclusion_threshold:
selectors.append(
DensitySelector(random_state=random.randint(1, 1e9)))
if random.random() > inclusion_threshold:
selectors.append(
OnionSelector(random_state=random.randint(1, 1e9)))
if not selectors:
selectors.append(
QuickShiftSelector(random_state=random.randint(1, 1e9)))
self.selector = CompositeSelector(selectors=selectors)
self.selector.randomize(data_matrix, amount=amount)
self.metric = 'rbf'
self.kwds = {'gamma': random.choice([10**x for x in range(-3, 3)])}
self.random_state = self.random_state ^ random.randint(1, 1e9)
class Projector(object):
"""Constructs features as the instance similarity to a set of instances as
defined by the selector.
Parameters
----------
selector : Selector
TODO.
scale : bool (default True)
If true then the data matrix returned is standardized to have 0 mean and unit variance
scaling_factor : float (default 0.8)
Multiplicative factor applied after normalization. This can be useful when data needs to be
post-processed by neural networks and one wishes to push data in a linear region.
random_state : int (deafault 1)
The seed used for the pseudo-random generator.
metric : string, or callable
The metric to use when calculating kernel between instances in a
feature array. If metric is a string, it must be one of the metrics
in pairwise.PAIRWISE_KERNEL_FUNCTIONS.
If metric is "precomputed", X is assumed to be a kernel matrix.
Alternatively, if metric is a callable function, it is called on each
pair of instances (rows) and the resulting value recorded. The callable
should take two arrays from X as input and return a value indicating
the distance between them.
**kwds : optional keyword parameters
Any further parameters are passed directly to the kernel function.
"""
def __init__(self, selector=AllSelector(),
scale=True,
scaling_factor=0.8,
random_state=1,
metric='rbf', **kwds):
self.selector = selector
self.scale = scale
self.scaling_factor = scaling_factor
self.scaler = StandardScaler()
self.metric = metric
self.kwds = kwds
self.random_state = random_state
def __repr__(self):
serial = []
serial.append('Projector:')
serial.append('metric: %s' % self.metric)
if self.kwds is None or len(self.kwds) == 0:
pass
else:
serial.append('params:')
serial.append(serialize_dict(self.kwds))
serial.append(str(self.selector))
return '\n'.join(serial)
def fit(self, data_matrix, target=None):
"""Fit the estimator on the samples.
Parameters
----------
data_matrix : array-like, shape = (n_samples, n_features)
Samples.
Returns
-------
self
"""
self.selected_instances = self.selector.fit_transform(data_matrix, target=target)
if self.scale:
self.scale = False
self.scaler.fit(self.transform(data_matrix))
self.scale = True
return self
def fit_transform(self, data_matrix, target=None):
"""Fit the estimator on the samples and transforms features as the instance
similarity to a set of instances as defined by the selector.
Parameters
----------
data_matrix : array, shape = (n_samples, n_features)
Samples.
target : TODO
Returns
-------
data_matrix : array, shape = (n_samples, n_features_new)
Transformed array.
"""
self.fit(data_matrix, target)
return self.transform(data_matrix)
def transform(self, data_matrix):
"""Transforms features as the instance similarity to a set of instances as
defined by the selector.
Parameters
----------
data_matrix : array, shape = (n_samples, n_features)
Samples.
Returns
-------
data_matrix : array, shape = (n_samples, n_features_new)
Transformed array.
"""
if self.selected_instances is None:
raise Exception('Error: attempt to use transform on non fit model')
if self.selected_instances.shape[0] == 0:
raise Exception('Error: attempt to use transform using 0 selectors')
# TODO: the first instance is more important than others in a selector, so it should
# receive a weight proportional to the rank e.g. 1/rank^p
# the selector should return also a rank information for each feature, note: for the
# composite selector it is important to distinguish the rank of multiple selectors
data_matrix_out = pairwise_kernels(data_matrix,
Y=self.selected_instances,
metric=self.metric,
**self.kwds)
if self.scale:
data_matrix_out = self.scaler.transform(data_matrix_out) * self.scaling_factor
return data_matrix_out
def randomize(self, data_matrix, amount=.5):
random.seed(self.random_state)
inclusion_threshold = random.uniform(amount, 1)
selectors = []
if random.random() > inclusion_threshold:
selectors.append(QuickShiftSelector(random_state=random.randint(1, 1e9)))
if random.random() > inclusion_threshold:
selectors.append(DecisionSurfaceSelector(random_state=random.randint(1, 1e9)))
if random.random() > inclusion_threshold:
selectors.append(SparseSelector(random_state=random.randint(1, 1e9)))
if random.random() > inclusion_threshold:
selectors.append(MaxVolSelector(random_state=random.randint(1, 1e9)))
if random.random() > inclusion_threshold:
selectors.append(DensitySelector(random_state=random.randint(1, 1e9)))
if random.random() > inclusion_threshold:
selectors.append(OnionSelector(random_state=random.randint(1, 1e9)))
if not selectors:
selectors.append(QuickShiftSelector(random_state=random.randint(1, 1e9)))
self.selector = CompositeSelector(selectors=selectors)
self.selector.randomize(data_matrix, amount=amount)
self.metric = 'rbf'
self.kwds = {'gamma': random.choice([10 ** x for x in range(-3, 3)])}
self.random_state = self.random_state ^ random.randint(1, 1e9)
