/usr/share/pyshared/mdp/nodes/scikits_nodes.py is in python-mdp 3.3-1.
This file is owned by root:root, with mode 0o644.
The actual contents of the file can be viewed below.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 | # -*- coding:utf-8; -*-
"""Wraps the algorithms defined in scikits.learn in MDP Nodes.
This module is based on the 0.6.X branch of scikits.learn .
"""
__docformat__ = "restructuredtext en"
try:
import sklearn
_sklearn_prefix = 'sklearn'
except ImportError:
import scikits.learn as sklearn
_sklearn_prefix = 'scikits.learn'
import inspect
import re
import mdp
class ScikitsException(mdp.NodeException):
"""Base class for exceptions in nodes wrapping scikits algorithms."""
pass
# import all submodules of sklearn (to work around lazy import)
from mdp.configuration import _version_too_old
if _version_too_old(sklearn.__version__, (0, 8)):
scikits_modules = ['ann', 'cluster', 'covariance', 'feature_extraction',
'feature_selection', 'features', 'gaussian_process', 'glm',
'linear_model', 'preprocessing', 'svm',
'pca', 'lda', 'hmm', 'fastica', 'grid_search', 'mixture',
'naive_bayes', 'neighbors', 'qda']
elif _version_too_old(sklearn.__version__, (0, 9)):
# package structure has been changed in 0.8
scikits_modules = ['svm', 'linear_model', 'naive_bayes', 'neighbors',
'mixture', 'hmm', 'cluster', 'decomposition', 'lda',
'covariance', 'cross_val', 'grid_search',
'feature_selection.rfe', 'feature_extraction.image',
'feature_extraction.text', 'pipelines', 'pls',
'gaussian_process', 'qda']
elif _version_too_old(sklearn.__version__, (0, 11)):
# from release 0.9 cross_val becomes cross_validation and hmm is deprecated
scikits_modules = ['svm', 'linear_model', 'naive_bayes', 'neighbors',
'mixture', 'cluster', 'decomposition', 'lda',
'covariance', 'cross_validation', 'grid_search',
'feature_selection.rfe', 'feature_extraction.image',
'feature_extraction.text', 'pipelines', 'pls',
'gaussian_process', 'qda', 'ensemble', 'manifold',
'metrics', 'preprocessing', 'tree']
else:
scikits_modules = ['svm', 'linear_model', 'naive_bayes', 'neighbors',
'mixture', 'cluster', 'decomposition', 'lda',
'covariance', 'cross_validation', 'grid_search',
'feature_selection', 'feature_extraction',
'pipeline', 'pls', 'gaussian_process', 'qda',
'ensemble', 'manifold', 'metrics', 'preprocessing',
'semi_supervised', 'tree', 'hmm']
for name in scikits_modules:
# not all modules may be available due to missing dependencies
# on the user system.
# we just ignore failing imports
try:
__import__(_sklearn_prefix + '.' + name)
except ImportError:
pass
_WS_LINE_RE = re.compile(r'^\s*$')
_WS_PREFIX_RE = re.compile(r'^(\s*)')
_HEADINGS_RE = re.compile(r'''^(Parameters|Attributes|Methods|Examples|Notes)\n
(----+|====+)''', re.M + re.X)
_UNDERLINE_RE = re.compile(r'----+|====+')
_VARWITHUNDER_RE = re.compile(r'(\s|^)([a-zA-Z_][a-zA-Z0-9_]*_)(\s|$|[,.])')
_HEADINGS = set(['Parameters', 'Attributes', 'Methods', 'Examples',
'Notes', 'References'])
_DOC_TEMPLATE = """
%s
This node has been automatically generated by wrapping the ``%s.%s`` class
from the ``sklearn`` library. The wrapped instance can be accessed
through the ``scikits_alg`` attribute.
%s
"""
def _gen_docstring(object, docsource=None):
module = object.__module__
name = object.__name__
if docsource is None:
docsource = object
docstring = docsource.__doc__
if docstring is None:
return None
lines = docstring.strip().split('\n')
for i,line in enumerate(lines):
if _WS_LINE_RE.match(line):
break
header = [line.strip() for line in lines[:i]]
therest = [line.rstrip() for line in lines[i+1:]]
body = []
if therest:
prefix = min(len(_WS_PREFIX_RE.match(line).group(1))
for line in therest if line)
quoteind = None
for i, line in enumerate(therest):
line = line[prefix:]
if line in _HEADINGS:
body.append('**%s**' % line)
elif _UNDERLINE_RE.match(line):
body.append('')
else:
line = _VARWITHUNDER_RE.sub(r'\1``\2``\3', line)
if quoteind:
if len(_WS_PREFIX_RE.match(line).group(1)) >= quoteind:
line = quoteind * ' ' + '- ' + line[quoteind:]
else:
quoteind = None
body.append('')
body.append(line)
if line.endswith(':'):
body.append('')
if i+1 < len(therest):
next = therest[i+1][prefix:]
quoteind = len(_WS_PREFIX_RE.match(next).group(1))
return _DOC_TEMPLATE % ('\n'.join(header), module, name, '\n'.join(body))
# TODO: generalize dtype support
# TODO: have a look at predict_proba for Classifier.prob
# TODO: inverse <-> generate/rvs
# TODO: deal with input_dim/output_dim
# TODO: change signature of overwritten functions
# TODO: wrap_scikits_instance
# TODO: add sklearn availability to test info strings
# TODO: which tests ? (test that particular algorithm are / are not trainable)
# XXX: if class defines n_components, allow output_dim, otherwise throw exception
# also for classifiers (overwrite _set_output_dim)
# Problem: sometimes they call it 'k' (e.g., algorithms in sklearn.cluster)
def apply_to_scikits_algorithms(current_module, action,
processed_modules=None,
processed_classes=None):
""" Function that traverses a module to find scikits algorithms.
'sklearn' algorithms are identified by the 'fit' 'predict',
or 'transform' methods. The 'action' function is applied to each found
algorithm.
action -- a function that is called with as action(class_), where
class_ is a class that defines the 'fit' or 'predict' method
"""
# only consider modules and classes once
if processed_modules is None:
processed_modules = []
if processed_classes is None:
processed_classes = []
if current_module in processed_modules:
return
processed_modules.append(current_module)
for member_name, member in current_module.__dict__.items():
if not member_name.startswith('_'):
# classes
if (inspect.isclass(member) and
member not in processed_classes):
processed_classes.append(member)
if ((hasattr(member, 'fit')
or hasattr(member, 'predict')
or hasattr(member, 'transform'))
and not member.__module__.endswith('_')):
action(member)
# other modules
elif (inspect.ismodule(member) and
member.__name__.startswith(_sklearn_prefix)):
apply_to_scikits_algorithms(member, action, processed_modules,
processed_classes)
return processed_classes
_OUTPUTDIM_ERROR = """'output_dim' keyword not supported.
Please set the output dimensionality using sklearn keyword
arguments (e.g., 'n_components', or 'k'). See the docstring of this
class for details."""
def wrap_scikits_classifier(scikits_class):
"""Wrap a sklearn classifier as an MDP Node subclass.
The wrapper maps these MDP methods to their sklearn equivalents:
- _stop_training -> fit
- _label -> predict
"""
newaxis = mdp.numx.newaxis
# create a wrapper class for a sklearn classifier
class ScikitsNode(mdp.ClassifierCumulator):
def __init__(self, input_dim=None, output_dim=None, dtype=None,
**kwargs):
if output_dim is not None:
# output_dim and n_components cannot be defined at the same time
if kwargs.has_key('n_components'):
msg = ("Dimensionality set both by "
"output_dim=%d and n_components=%d""")
raise ScikitsException(msg % (output_dim,
kwargs['n_components']))
super(ScikitsNode, self).__init__(input_dim=input_dim,
output_dim=output_dim,
dtype=dtype)
self.scikits_alg = scikits_class(**kwargs)
# ---- re-direct training and execution to the wrapped algorithm
def _stop_training(self, **kwargs):
super(ScikitsNode, self)._stop_training(self)
return self.scikits_alg.fit(self.data, self.labels, **kwargs)
def _label(self, x):
return self.scikits_alg.predict(x)[:, newaxis]
# ---- administrative details
@staticmethod
def is_invertible():
return False
@staticmethod
def is_trainable():
"""Return True if the node can be trained, False otherwise."""
return hasattr(scikits_class, 'fit')
# NOTE: at this point scikits nodes can only support up to
# 64-bits floats because some call numpy.linalg.svd, which for
# some reason does not support higher precisions
def _get_supported_dtypes(self):
"""Return the list of dtypes supported by this node.
The types can be specified in any format allowed by numpy.dtype."""
return ['float32', 'float64']
# modify class name and docstring
ScikitsNode.__name__ = scikits_class.__name__ + 'ScikitsLearnNode'
ScikitsNode.__doc__ = _gen_docstring(scikits_class)
# change the docstring of the methods to match the ones in sklearn
# methods_dict maps ScikitsNode method names to sklearn method names
methods_dict = {'__init__': '__init__',
'stop_training': 'fit',
'label': 'predict'}
if hasattr(scikits_class, 'predict_proba'):
methods_dict['prob'] = 'predict_proba'
for mdp_name, scikits_name in methods_dict.items():
mdp_method = getattr(ScikitsNode, mdp_name)
scikits_method = getattr(scikits_class, scikits_name)
if hasattr(scikits_method, 'im_func'):
# some scikits algorithms do not define an __init__ method
# the one inherited from 'object' is a
# "<slot wrapper '__init__' of 'object' objects>"
# which does not have a 'im_func' attribute
mdp_method.im_func.__doc__ = _gen_docstring(scikits_class,
scikits_method.im_func)
if scikits_class.__init__.__doc__ is None:
ScikitsNode.__init__.im_func.__doc__ = _gen_docstring(scikits_class)
return ScikitsNode
def wrap_scikits_transformer(scikits_class):
"""Wrap a sklearn transformer as an MDP Node subclass.
The wrapper maps these MDP methods to their sklearn equivalents:
_stop_training -> fit
_execute -> transform
"""
# create a wrapper class for a sklearn transformer
class ScikitsNode(mdp.Cumulator):
def __init__(self, input_dim=None, output_dim=None, dtype=None, **kwargs):
if output_dim is not None:
raise ScikitsException(_OUTPUTDIM_ERROR)
super(ScikitsNode, self).__init__(input_dim=input_dim,
output_dim=output_dim,
dtype=dtype)
self.scikits_alg = scikits_class(**kwargs)
# ---- re-direct training and execution to the wrapped algorithm
def _stop_training(self, **kwargs):
super(ScikitsNode, self)._stop_training(self)
return self.scikits_alg.fit(self.data, **kwargs)
def _execute(self, x):
return self.scikits_alg.transform(x)
# ---- administrative details
@staticmethod
def is_invertible():
return False
@staticmethod
def is_trainable():
"""Return True if the node can be trained, False otherwise."""
return hasattr(scikits_class, 'fit')
# NOTE: at this point scikits nodes can only support up to
# 64-bits floats because some call numpy.linalg.svd, which for
# some reason does not support higher precisions
def _get_supported_dtypes(self):
"""Return the list of dtypes supported by this node.
The types can be specified in any format allowed by numpy.dtype."""
return ['float32', 'float64']
# modify class name and docstring
ScikitsNode.__name__ = scikits_class.__name__ + 'ScikitsLearnNode'
ScikitsNode.__doc__ = _gen_docstring(scikits_class)
# change the docstring of the methods to match the ones in sklearn
# methods_dict maps ScikitsNode method names to sklearn method names
methods_dict = {'__init__': '__init__',
'stop_training': 'fit',
'execute': 'transform'}
for mdp_name, scikits_name in methods_dict.items():
mdp_method = getattr(ScikitsNode, mdp_name)
scikits_method = getattr(scikits_class, scikits_name, None)
if hasattr(scikits_method, 'im_func'):
# some scikits algorithms do not define an __init__ method
# the one inherited from 'object' is a
# "<slot wrapper '__init__' of 'object' objects>"
# which does not have a 'im_func' attribute
mdp_method.im_func.__doc__ = _gen_docstring(scikits_class,
scikits_method.im_func)
if scikits_class.__init__.__doc__ is None:
ScikitsNode.__init__.im_func.__doc__ = _gen_docstring(scikits_class)
return ScikitsNode
def wrap_scikits_predictor(scikits_class):
"""Wrap a sklearn transformer as an MDP Node subclass.
The wrapper maps these MDP methods to their sklearn equivalents:
_stop_training -> fit
_execute -> predict
"""
# create a wrapper class for a sklearn predictor
class ScikitsNode(mdp.Cumulator):
def __init__(self, input_dim=None, output_dim=None, dtype=None, **kwargs):
if output_dim is not None:
raise ScikitsException(_OUTPUTDIM_ERROR)
super(ScikitsNode, self).__init__(input_dim=input_dim,
output_dim=output_dim,
dtype=dtype)
self.scikits_alg = scikits_class(**kwargs)
# ---- re-direct training and execution to the wrapped algorithm
def _stop_training(self, **kwargs):
super(ScikitsNode, self)._stop_training(self)
return self.scikits_alg.fit(self.data, **kwargs)
def _execute(self, x):
return self.scikits_alg.predict(x)
# ---- administrative details
@staticmethod
def is_invertible():
return False
@staticmethod
def is_trainable():
"""Return True if the node can be trained, False otherwise."""
return hasattr(scikits_class, 'fit')
# NOTE: at this point scikits nodes can only support up to 64-bits floats
# because some call numpy.linalg.svd, which for some reason does not
# support higher precisions
def _get_supported_dtypes(self):
"""Return the list of dtypes supported by this node.
The types can be specified in any format allowed by numpy.dtype."""
return ['float32', 'float64']
# modify class name and docstring
ScikitsNode.__name__ = scikits_class.__name__ + 'ScikitsLearnNode'
ScikitsNode.__doc__ = _gen_docstring(scikits_class)
# change the docstring of the methods to match the ones in sklearn
# methods_dict maps ScikitsNode method names to sklearn method names
methods_dict = {'__init__': '__init__',
'stop_training': 'fit',
'execute': 'predict'}
for mdp_name, scikits_name in methods_dict.items():
mdp_method = getattr(ScikitsNode, mdp_name)
scikits_method = getattr(scikits_class, scikits_name)
if hasattr(scikits_method, 'im_func'):
# some scikits algorithms do not define an __init__ method
# the one inherited from 'object' is a
# "<slot wrapper '__init__' of 'object' objects>"
# which does not have a 'im_func' attribute
mdp_method.im_func.__doc__ = _gen_docstring(scikits_class,
scikits_method.im_func)
if scikits_class.__init__.__doc__ is None:
ScikitsNode.__init__.im_func.__doc__ = _gen_docstring(scikits_class)
return ScikitsNode
#list candidate nodes
def print_public_members(class_):
"""Print methods of sklearn algorithm.
"""
print '\n', '-' * 15
print '%s (%s)' % (class_.__name__, class_.__module__)
for attr_name in dir(class_):
attr = getattr(class_, attr_name)
#print attr_name, type(attr)
if not attr_name.startswith('_') and inspect.ismethod(attr):
print ' -', attr_name
#apply_to_scikits_algorithms(sklearn, print_public_members)
def wrap_scikits_algorithms(scikits_class, nodes_list):
"""NEED DOCSTRING."""
name = scikits_class.__name__
if (name[:4] == 'Base' or name == 'LinearModel'
or name.startswith('EllipticEnvelop')
or name.startswith('ForestClassifier')):
return
if issubclass(scikits_class, sklearn.base.ClassifierMixin) and \
hasattr(scikits_class, 'fit'):
nodes_list.append(wrap_scikits_classifier(scikits_class))
# Some (abstract) transformers do not implement fit.
elif hasattr(scikits_class, 'transform') and hasattr(scikits_class, 'fit'):
nodes_list.append(wrap_scikits_transformer(scikits_class))
elif hasattr(scikits_class, 'predict') and hasattr(scikits_class, 'fit'):
nodes_list.append(wrap_scikits_predictor(scikits_class))
scikits_nodes = []
apply_to_scikits_algorithms(sklearn,
lambda c: wrap_scikits_algorithms(c, scikits_nodes))
# add scikit nodes to dictionary
#scikits_module = new.module('scikits')
DICT_ = {}
for wrapped_c in scikits_nodes:
#print wrapped_c.__name__
DICT_[wrapped_c.__name__] = wrapped_c
|