def setData(self,x,y,*args,**kwargin):
"""set Data
x: inputs [N,D]
t: targets [N]
- targets are either -1,+1 or False/True
"""
assert isinstance(y,S.ndarray), 'setData requires numpy arrays'
#check whether t is bool
if y.dtype=='bool':
y_ = S.ones([y.shape[0]])
y_[y] = +1
y_[~y] = -1
y = y_
else:
assert len(SP.unique(y))==2, 'need either binary inputs or inputs of length 2 for classification'
GPEP.setData(self,x=x,y=y,*args,**kwargin)
def setData(self, x, y, *args, **kwargin):
"""set Data
x: inputs [N,D]
t: targets [N]
- targets are either -1,+1 or False/True
"""
assert isinstance(y, S.ndarray), 'setData requires numpy arrays'
#check whether t is bool
if y.dtype == 'bool':
y_ = S.ones([y.shape[0]])
y_[y] = +1
y_[~y] = -1
y = y_
else:
assert len(
SP.unique(y)
) == 2, 'need either binary inputs or inputs of length 2 for classification'
GPEP.setData(self, x=x, y=y, *args, **kwargin)
def predict(self,*argin,**kwargin):
"""Binary classification prediction"""
#1. get Gaussian prediction
[MU,S2] = GPEP.predict(self,*argin,**kwargin)
#2. push thorugh sigmoid
#predictive distribution is int_-inf^+inf normal(f|mu,s2)sigmoid(f)
Pt = sigmoid ( MU / S.sqrt(1+S2))
return [Pt,MU,S2]
pass
def predict(self, *argin, **kwargin):
"""Binary classification prediction"""
#1. get Gaussian prediction
[MU, S2] = GPEP.predict(self, *argin, **kwargin)
#2. push thorugh sigmoid
#predictive distribution is int_-inf^+inf normal(f|mu,s2)sigmoid(f)
Pt = sigmoid(MU / S.sqrt(1 + S2))
return [Pt, MU, S2]
pass