class StatModel:
def __init__(self, k):
"""
We approximate by the mean of our nearest neighbors
"""
self.k=k
self.idx=PseudoIncrementalIndex(SashIndex)
def dump(self,file_):
assert(False)
@staticmethod
def load(file_, *args, **kwargs):
assert(False)
def train(self,A,B,online=False):
## basically train consists in creating 1 or 2 sahs
assert(A!=None)
assert(B!=None)
self.idx.add_many(A,B)
self.idx.recompute()
def predict(self,A,AN=None,B=None,BN=None,log=False):
"""
Predict what will be the label according to $k$-closest neighbors
"""
### basically train consists in checking out the average distance of neighbors
### basically train consists in checking out the average distance of neighbors compared to oponents features...
#
# basically we consider density estimation as a
#
res=self.idx.getitems(A,self.k)
return map(lambda x:x[0][0][0],res)
#parzenestimator()
#return scipy.mean(map(self.idx.getitem(obs,self.k),lambda x:x[1]))
def memory_cost(self, *args, **kwargs):
assert(False)
def cpu_cost(self, *args, **kwargs):
assert(False)
def random_improve(self,value,amount=0.5, prec=1):
## TRY TO RANDOMLY CHANGE WITH A NEAREST NEIGHBOR
#print "random_improve not yet implemented doing sample instead !!! hist"
#return self.sample(value.shape[0])
assert(False)
def __init__(self, k):
"""
We approximate by the mean of our nearest neighbors
"""
self.k=k
self.idx=PseudoIncrementalIndex(SashIndex)
