def _get_statistics(self,data,D):
n = getdatasize(data)
if n > 0:
if isinstance(data,np.ndarray):
Syy = data[:,-D:].T.dot(data[:,-D:])
Sytyt = data[:,:-D].T.dot(data[:,:-D])
Syyt = data[:,-D:].T.dot(data[:,:-D])
if self.affine:
Syyt = np.hstack((data[:,-D:].sum(0)[:,na],Syyt))
Sytytsum = data[:,:-D].sum(0)
Sytyt = np.vstack((
np.hstack(((n,),Sytytsum)),
np.hstack((Sytytsum[:,na],Sytyt))
))
else:
Syy = sum(d[:,-D:].T.dot(d[:,-D:]) for d in data)
Sytyt = sum(d[:,:-D].T.dot(d[:,:-D]) for d in data)
Syyt = sum(d[:,-D:].T.dot(d[:,:-D]) for d in data)
if self.affine:
Syyt = np.hstack((sum(d[:,-D:].sum(0) for d in data)[:,na],Syyt))
Sytytsum = sum(d[:,:-D].sum(0) for d in data)
Sytyt = np.vstack((
np.hstack(((n,),Sytytsum)),
np.hstack((Sytytsum[:,na],Sytyt))
))
else:
Syy = Sytyt = Syyt = None
return Syy,Sytyt,Syyt,n
def _get_statistics(self,data,D):
# NOTE: similar to pybasicbayes/distributions.py:GaussianFixedMean
n = getdatasize(data)
if n > 0:
if isinstance(data,np.ndarray):
centered = self._center_data(data)
sumsq = centered.T.dot(centered)
else:
sumsq = sum(c.T.dot(c) for c in map(self._center_data,data))
else:
sumsq = None
return n, sumsq
def resample(self,data,niter=25,temp=None):
# doesn't keep a reference to the data like a model would
assert isinstance(data,list) or isinstance(data,np.ndarray)
if getdatasize(data) > 0:
if not isinstance(data,np.ndarray):
data = np.concatenate(data)
self.add_data(data)
for itr in range(niter):
self.resample_model(temp=temp)
self.labels_list.pop()
else:
self.resample_model(temp=temp)
def resample(self, data, niter=25, temp=None):
# doesn't keep a reference to the data like a model would
assert isinstance(data, list) or isinstance(data, np.ndarray)
if getdatasize(data) > 0:
if not isinstance(data, np.ndarray):
data = np.concatenate(data)
self.add_data(data)
for itr in range(niter):
self.resample_model(temp=temp)
self.labels_list.pop()
else:
self.resample_model(temp=temp)
def max_likelihood(self,data,weights=None):
if weights is not None:
raise NotImplementedError
assert isinstance(data,list) or isinstance(data,np.ndarray)
if isinstance(data,np.ndarray):
data = [np.asarray(data,dtype=np.float64)]
else:
data = map(lambda x: np.asarray(x,dtype=np.float64), data)
if getdatasize(data) > 0:
for d in data:
self.add_data(d)
prev_like = sum(self.log_likelihood(d).sum() for d in data)
for itr in range(100):
self.EM_step()
new_like = sum(self.log_likelihood(d).sum() for d in data)
if new_like <= prev_like + 0.1:
break
else:
prev_like = new_like
for d in data:
self.labels_list.pop()
def max_likelihood(self, data, weights=None):
if weights is not None:
raise NotImplementedError
assert isinstance(data, list) or isinstance(data, np.ndarray)
if isinstance(data, np.ndarray):
data = [np.asarray(data, dtype=np.float64)]
else:
data = map(lambda x: np.asarray(x, dtype=np.float64), data)
if getdatasize(data) > 0:
for d in data:
self.add_data(d)
prev_like = sum(self.log_likelihood(d).sum() for d in data)
for itr in range(100):
self.EM_step()
new_like = sum(self.log_likelihood(d).sum() for d in data)
if new_like <= prev_like + 0.1:
break
else:
prev_like = new_like
for d in data:
self.labels_list.pop()