def __init__(self, c00=True, c0t=False, ctt=False, remove_constant_mean=None, remove_data_mean=False, reversible=False,
bessel=True, sparse_mode='auto', modify_data=False, lag=0, weights=None, stride=1, skip=0,
chunksize=NotImplemented, ncov_max=float('inf'), column_selection=None, diag_only=False):
super(LaggedCovariance, self).__init__()
if chunksize is not NotImplemented:
import warnings
from pyemma.util.exceptions import PyEMMA_DeprecationWarning
warnings.warn('passed deprecated argument chunksize to LaggedCovariance. Will be ignored!',
category=PyEMMA_DeprecationWarning)
if (c0t or ctt) and lag == 0:
raise ValueError("lag must be positive if c0t=True or ctt=True")
if remove_constant_mean is not None and remove_data_mean:
raise ValueError('Subtracting the data mean and a constant vector simultaneously is not supported.')
if remove_constant_mean is not None:
remove_constant_mean = ensure_float_vector(remove_constant_mean)
if column_selection is not None and diag_only:
raise ValueError('Computing only parts of the diagonal is not supported.')
if diag_only and sparse_mode is not 'dense':
if sparse_mode is 'sparse':
self.logger.warning('Computing diagonal entries only is not implemented for sparse mode. Switching to dense mode.')
sparse_mode = 'dense'
self.set_params(c00=c00, c0t=c0t, ctt=ctt, remove_constant_mean=remove_constant_mean,
remove_data_mean=remove_data_mean, reversible=reversible,
sparse_mode=sparse_mode, modify_data=modify_data, lag=lag,
bessel=bessel,
weights=weights, stride=stride, skip=skip, ncov_max=ncov_max,
column_selection=column_selection, diag_only=diag_only)
self._rc = None
def __init__(self,
c00=True,
c0t=False,
ctt=False,
remove_constant_mean=None,
remove_data_mean=False,
reversible=False,
bessel=True,
sparse_mode='auto',
modify_data=False,
lag=0,
weights=None,
stride=1,
skip=0,
chunksize=None,
ncov_max=float('inf')):
super(LaggedCovariance, self).__init__(chunksize=chunksize)
if (c0t or ctt) and lag == 0:
raise ValueError("lag must be positive if c0t=True or ctt=True")
if is_float_vector(weights):
weights = ensure_float_vector(weights)
if remove_constant_mean is not None and remove_data_mean:
raise ValueError(
'Subtracting the data mean and a constant vector simultaneously is not supported.'
)
if remove_constant_mean is not None:
remove_constant_mean = ensure_float_vector(remove_constant_mean)
self.set_params(c00=c00,
c0t=c0t,
ctt=ctt,
remove_constant_mean=remove_constant_mean,
remove_data_mean=remove_data_mean,
reversible=reversible,
sparse_mode=sparse_mode,
modify_data=modify_data,
lag=lag,
bessel=bessel,
weights=weights,
stride=stride,
skip=skip,
ncov_max=ncov_max)
self._rc = None
self._used_data = 0
def expectation(self, a):
a = _types.ensure_float_vector(a, require_order=True)
# are we on microstates space?
if len(a) == self.nstates_obs:
# project to hidden and compute
a = _np.dot(self.observation_probabilities, a)
# now we are on macrostate space, or something is wrong
if len(a) == self.nstates:
return _MSM.expectation(self, a)
else:
raise ValueError(
'observable vector a has size %s which is incompatible with both hidden (%s) '
'and observed states (%s)' %
(len(a), self.nstates, self.nstates_obs))