class PerceptronBinary( Binary ):
""" Linear classifier with online learning and perceptron update
(in primal form)"""
def __init__( self, bias=False ):
Binary.__init__(self, bias=bias )
self._learner = OnlineLearner(self)
return
def learn( self, data, epochs=1, start_iter=0, write_every_iter=False, modelpath=None, forget_first=False):
instances = self._get_train_stream( data )
self._learner.learn( instances, epochs, start_iter, write_every_iter, modelpath, forget_first )
return
def update( self, instance, prediction ):
""" perceptron update rule:
w = w + y * x
"""
w = self._weights
fv = instance.get_fv()
t_lab = int(instance.get_target_label())
p_lab = int(prediction.get_label())
error = (p_lab != t_lab)
# ...
if (error) :
self._weights = w + dot(t_lab, fv)
return error
class PerceptronBinary( Binary ): """ Linear classifier with online learning and perceptron update (in primal form)""" def __init__( self, bias=False ): Binary.__init__(self, bias=bias ) self._learner = OnlineLearner(self) return def learn( self, data, epochs=1, start_iter=0, write_every_iter=False, modelpath=None, forget_first=False): instances = self._get_train_stream( data ) self._learner.learn( instances, epochs, start_iter, write_every_iter, modelpath, forget_first ) return def update( self, instance, prediction ): """ perceptron update rule: NOOOOOP!!! -> w = w + (y - y^) * x """ w = self._weights # x_t fv = instance.get_fv() # y_t t_lab = int(instance.get_target_label()) # predict y_t p_lab = int(prediction.get_label()) error = (p_lab != t_lab) if error: self._weights += t_lab * fv return error
class PerceptronABinary(Binary):
""" Linear classifier with online learning and perceptron update
(in primal form)"""
def __init__(self, bias=False):
Binary.__init__(self, bias=bias)
self._learner = OnlineLearner(self)
self._w_average = self._weights
def _decode(self, instance, weights):
""" return prediction in {-1, 1}
for current instance based on linear combination of given
weight parameters """
fv = instance.get_fv()
score = dot(self._w_average, fv)
return Prediction(score)
def resize_weights(self, instance):
if len(self._weights) != self._feature_alphabet.size():
self._weights.resize(self._feature_alphabet.size())
self._w_average.resize(self._feature_alphabet.size())
return
def learn(self,
data,
epochs=1,
start_iter=0,
write_every_iter=False,
modelpath=None,
forget_first=False):
instances = self._get_train_stream(data)
self._learner.learn(instances, epochs, start_iter, write_every_iter,
modelpath, forget_first)
return
def update(self, instance, prediction):
""" perceptron update rule:
NOOOOOP!!! -> w = w + (y - y^) * x
"""
w = self._weights
self._w_average += w
# x_t
fv = instance.get_fv()
# y_t
t_lab = int(instance.get_target_label())
# predict y_t
p_lab = int(prediction.get_label())
error = (p_lab != t_lab)
if error:
self._weights += t_lab * fv
return error
class PerceptronABinary( Binary ):
""" Linear classifier with online learning and perceptron update
(in primal form)"""
def __init__( self, bias=False ):
Binary.__init__(self, bias=bias )
self._learner = OnlineLearner(self)
self._w_average = self._weights
def _decode( self, instance, weights ):
""" return prediction in {-1, 1}
for current instance based on linear combination of given
weight parameters """
fv = instance.get_fv()
score = dot( self._w_average, fv )
return Prediction( score )
def resize_weights(self, instance):
if len(self._weights) != self._feature_alphabet.size():
self._weights.resize(self._feature_alphabet.size())
self._w_average.resize(self._feature_alphabet.size())
return
def learn( self, data, epochs=1, start_iter=0, write_every_iter=False, modelpath=None, forget_first=False):
instances = self._get_train_stream( data )
self._learner.learn( instances, epochs, start_iter, write_every_iter, modelpath, forget_first )
return
def update( self, instance, prediction ):
""" perceptron update rule:
NOOOOOP!!! -> w = w + (y - y^) * x
"""
w = self._weights
self._w_average += w
# x_t
fv = instance.get_fv()
# y_t
t_lab = int(instance.get_target_label())
# predict y_t
p_lab = int(prediction.get_label())
error = (p_lab != t_lab)
if error:
self._weights += t_lab * fv
return error
class PerceptronBinary(Binary):
""" Linear classifier with online learning and perceptron update
(in primal form)"""
def __init__(self, bias=False):
Binary.__init__(self, bias=bias)
self._learner = OnlineLearner(self)
return
def learn(self,
data,
epochs=1,
start_iter=0,
write_every_iter=False,
modelpath=None,
forget_first=False):
instances = self._get_train_stream(data)
self._learner.learn(instances, epochs, start_iter, write_every_iter,
modelpath, forget_first)
return
def update(self, instance, prediction):
""" perceptron update rule:
NOOOOOP!!! -> w = w + (y - y^) * x
"""
w = self._weights
# x_t
fv = instance.get_fv()
# y_t
t_lab = int(instance.get_target_label())
# predict y_t
p_lab = int(prediction.get_label())
error = (p_lab != t_lab)
if error:
self._weights += t_lab * fv
return error
def __init__(self, bias=False):
Binary.__init__(self, bias=bias)
self._learner = OnlineLearner(self)
self._w_average = self._weights
def __init__(self, bias=False):
Binary.__init__(self, bias=bias)
self._learner = OnlineLearner(self)
return
def __init__( self, bias=False ): Binary.__init__(self, bias=bias ) self._learner = OnlineLearner(self) return
def __init__( self, bias=False ): Binary.__init__(self, bias=bias ) self._learner = OnlineLearner(self) self._w_average = self._weights
def __init__( self, bias=False, avgmode=False ): Binary.__init__(self, bias=bias, avgmode=avgmode ) self._learner = OnlineLearner(self) return
