def set_model_params(self,
samples=None,
conf=0.95,
P=None,
pi=None,
reversible=None,
dt_model='1 step',
neig=None):
"""
Parameters
----------
samples : list of MSM objects
sampled MSMs
conf : float, optional, default=0.68
Confidence interval. By default one-sigma (68.3%) is used. Use 95.4% for two sigma or 99.7% for three sigma.
"""
# set model parameters of superclass
SampledModel.set_model_params(self, samples=samples, conf=conf)
MSM.set_model_params(self,
P=P,
pi=pi,
reversible=reversible,
dt_model=dt_model,
neig=neig)
def set_model_params(self,
P=None,
pobs=None,
pi=None,
reversible=None,
dt_model='1 step',
neig=None):
"""
Parameters
----------
P : ndarray(m,m)
coarse-grained or hidden transition matrix
pobs : ndarray (m,n)
observation probability matrix from hidden to observable discrete
states
pi : ndarray(m), optional, default=None
stationary or distribution. Can be optionally given in case if
it was already computed, e.g. by the estimator.
reversible : bool, optional, default=None
whether P is reversible with respect to its stationary distribution.
If None (default), will be determined from P
dt_model : str, optional, default='1 step'
time step of the model
neig : int or None
The number of eigenvalues / eigenvectors to be kept. If set to
None, all eigenvalues will be used.
"""
_MSM.set_model_params(self,
P=P,
pi=pi,
reversible=reversible,
dt_model=dt_model,
neig=neig)
# set P and derived quantities if available
if pobs is not None:
# check and save copy of output probability
assert _types.is_float_matrix(
pobs), 'pobs is not a matrix of floating numbers'
assert _np.allclose(pobs.sum(axis=1),
1), 'pobs is not a stochastic matrix'
self._nstates_obs = pobs.shape[1]
self.update_model_params(pobs=pobs)
def __init__(self, P, pobs, pi=None, dt_model='1 step'):
"""
Parameters
----------
Pcoarse : ndarray (m,m)
coarse-grained or hidden transition matrix
Pobs : ndarray (m,n)
observation probability matrix from hidden to observable discrete states
dt_model : str, optional, default='1 step'
time step of the model
"""
# construct superclass and check input
_MSM.__init__(self, P, dt_model=dt_model)
self.set_model_params(pobs=pobs, pi=pi, dt_model=dt_model)
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))
def __init__(self, samples, ref=None, conf=0.95):
r""" Constructs a sampled MSM
Parameters
----------
samples : list of MSM
Sampled MSM objects
ref : obj of type :class:`pyemma.msm.MaximumLikelihoodMSM` or :class:`pyemma.msm.BayesianMSM`
Single-point estimator, e.g. containing a maximum likelihood or mean MSM
conf : float, optional, default=0.95
Confidence interval. By default two-sigma (95.4%) is used. Use 95.4% for two sigma or 99.7% for three sigma.
"""
# validate input
assert is_iterable(
samples), 'samples must be a list of MSM objects, but is not.'
assert isinstance(
samples[0],
MSM), 'samples must be a list of MSM objects, but is not.'
# construct superclass 1
SampledModel.__init__(self, samples, conf=conf)
# construct superclass 2
if ref is None:
Pref = self.sample_mean('P')
MSM.__init__(self,
Pref,
dt_model=samples[0].dt_model,
neig=samples[0].neig,
ncv=samples[0].ncv)
else:
MSM.__init__(self,
ref.Pref,
pi=ref.pi,
reversible=ref.reversible,
dt_model=ref.dt_model,
neig=ref.neig,
ncv=ref.ncv)
def fingerprint_relaxation(self, p0, a, k=None, ncv=None):
# basic checks for a and b
p0 = _types.ensure_ndarray(p0, ndim=1, kind='numeric')
a = _types.ensure_ndarray(a, ndim=1, kind='numeric', size=len(p0))
# are we on microstates space?
if len(a) == self.nstates_obs:
p0 = _np.dot(self.observation_probabilities, p0)
a = _np.dot(self.observation_probabilities, a)
# now we are on macrostate space, or something is wrong
if len(a) == self.nstates:
return _MSM.fingerprint_relaxation(self, p0, a)
else:
raise ValueError(
'observable vectors have size %s which is incompatible with both hidden (%s)'
' and observed states (%s)' %
(len(a), self.nstates, self.nstates_obs))
def correlation(self, a, b=None, maxtime=None, k=None, ncv=None):
# basic checks for a and b
a = _types.ensure_ndarray(a, ndim=1, kind='numeric')
b = _types.ensure_ndarray_or_None(b,
ndim=1,
kind='numeric',
size=len(a))
# are we on microstates space?
if len(a) == self.nstates_obs:
a = _np.dot(self.observation_probabilities, a)
if b is not None:
b = _np.dot(self.observation_probabilities, b)
# now we are on macrostate space, or something is wrong
if len(a) == self.nstates:
return _MSM.correlation(self, a, b=b, maxtime=maxtime)
else:
raise ValueError(
'observable vectors have size %s which is incompatible with both hidden (%s)'
' and observed states (%s)' %
(len(a), self.nstates, self.nstates_obs))
def set_model_params(self,
samples=None,
conf=0.95,
P=None,
pi=None,
reversible=None,
dt_model='1 step',
neig=None):
"""
Parameters
----------
samples : list of MSM objects
sampled MSMs
conf : float, optional, default=0.68
Confidence interval. By default one-sigma (68.3%) is used. Use 95.4% for two sigma or 99.7% for three sigma.
"""
# set model parameters of superclass
SampledModel.set_model_params(self, samples=samples, conf=conf)
MSM.set_model_params(self,
P=P,
pi=pi,
reversible=reversible,
dt_model=dt_model,
neig=neig)
#
# class SampledEstimatedMSM(EstimatedMSM, SampledModel):
#
# def __init__(self, samples, ref, Pref='mle', conf=0.95):
# r""" Constructs a sampled MSM
#
# Parameters
# ----------
# samples : list of MSM
# Sampled MSM objects
# ref : EstimatedMSM
# Single-point estimator, e.g. containing a maximum likelihood or mean MSM
# conf : float, optional, default=0.68
# Confidence interval. By default one-sigma (68.3%) is used. Use 95.4% for two sigma or 99.7% for three sigma.
#
# """
# # construct superclass 1
# SampledModel.__init__(self, samples, conf=conf)
# # use reference or mean MSM.
# if ref is None:
# Pref = self.sample_mean('P')
# else:
# Pref = ref.P
# # construct superclass 2
# EstimatedMSM.__init__(self, ref.discrete_trajectories_full, ref.timestep, ref.lagtime, ref.connectivity,
# ref.active_set, ref.connected_sets, ref.count_matrix_full, ref.count_matrix_active, Pref)
# def _do_sample_eigendecomposition(self, k, ncv=None):
# """Conducts the eigenvalue decompositions for all sampled matrices.
#
# Stores k eigenvalues, left and right eigenvectors for all sampled matrices
#
# Parameters
# ----------
# k : int
# The number of eigenvalues / eigenvectors to be kept
# ncv : int (optional)
# Relevant for eigenvalue decomposition of reversible transition matrices.
# ncv is the number of Lanczos vectors generated, `ncv` must be greater than k;
# it is recommended that ncv > 2*k
#
# """
# from msmtools.analysis import rdl_decomposition
# from pyemma.util import linalg
#
# # left eigenvectors
# self.sample_Ls = np.empty((self._nsample), dtype=object)
# # eigenvalues
# self.sample_eigenvalues = np.empty((self._nsample), dtype=object)
# # right eigenvectors
# self.sample_Rs = np.empty((self._nsample), dtype=object)
# # eigenvector assignments
# self.sample_eig_assignments = np.empty((self._nsample), dtype=object)
#
# for i in range(self._nsample):
# if self._reversible:
# R, D, L = rdl_decomposition(self.sample_Ps[i], k=k, norm='reversible', ncv=ncv)
# # everything must be real-valued
# R = R.real
# D = D.real
# L = L.real
# else:
# R, D, L = rdl_decomposition(self.sample_Ps[i], k=k, norm='standard', ncv=ncv)
# # assign ordered
# I = linalg.match_eigenvectors(self.eigenvectors_right(), R,
# w_ref=self.stationary_distribution, w=self.sample_mus[i])
# self.sample_Ls[i] = L[I,:]
# self.sample_eigenvalues[i] = np.diag(D)[I]
# self.sample_Rs[i] = R[:,I]
#
# def _ensure_sample_eigendecomposition(self, k=None, ncv=None):
# """Ensures that eigendecomposition has been performed with at least k eigenpairs
#
# k : int
# number of eigenpairs needed. This setting is mandatory for sparse transition matrices
# (if you set sparse=True in the initialization). For dense matrices, k will be ignored
# as all eigenvalues and eigenvectors will be computed and stored.
# ncv : int (optional)
# Relevant for eigenvalue decomposition of reversible transition matrices.
# ncv is the number of Lanczos vectors generated, `ncv` must be greater than k;
# it is recommended that ncv > 2*k
#
# """
# # check input?
# if self._sparse:
# if k is None:
# raise ValueError(
# 'You have requested sparse=True, then the number of eigenvalues neig must also be set.')
# else:
# # override setting - we anyway have to compute all eigenvalues, so we'll also store them.
# k = self._nstates
# # ensure that eigenvalue decomposition with k components is done.
# try:
# m = len(self.sample_eigenvalues[0]) # this will raise and exception if self._eigenvalues doesn't exist yet.
# if m < k:
# # not enough eigenpairs present - recompute:
# self._do_sample_eigendecomposition(k, ncv=ncv)
# except:
# # no eigendecomposition yet - compute:
# self._do_sample_eigendecomposition(k, ncv=ncv)
#
# @property
# def stationary_distribution_mean(self):
# """Sample mean for the stationary distribution on the active set.
#
# See also
# --------
# MSM.stationary_distribution
#
# """
# return np.mean(self.sample_mus, axis=0)
#
# @property
# def stationary_distribution_std(self):
# """Sample standard deviation for the stationary distribution on the active set.
#
# See also
# --------
# MSM.stationary_distribution
#
# """
# return np.std(self.sample_mus, axis=0)
#
# @property
# def stationary_distribution_conf(self):
# """Sample confidence interval for the stationary distribution on the active set.
#
# See also
# --------
# MSM.stationary_distribution
#
# """
# return stat.confidence_interval(self.sample_mus, alpha=self._confidence)
#
# def eigenvalues_mean(self, k=None, ncv=None):
# """Sample mean for the eigenvalues.
#
# See also
# --------
# MSM.eigenvalues
#
# """
# self._ensure_sample_eigendecomposition(k=k, ncv=ncv)
# return np.mean(self.sample_eigenvalues, axis=0)
#
# def eigenvalues_std(self, k=None, ncv=None):
# """Sample standard deviation for the eigenvalues.
#
# See also
# --------
# MSM.eigenvalues
#
# """
# self._ensure_sample_eigendecomposition(k=k, ncv=ncv)
# return np.std(self.sample_eigenvalues, axis=0)
#
# def eigenvalues_conf(self, k=None, ncv=None):
# """Sample confidence interval for the eigenvalues.
#
# See also
# --------
# MSM.eigenvalues
#
# """
# self._ensure_sample_eigendecomposition(k=k, ncv=ncv)
# return stat.confidence_interval(self.sample_eigenvalues, alpha=self._confidence)
#
# def eigenvectors_left_mean(self, k=None, ncv=None):
# """Sample mean for the left eigenvectors.
#
# See also
# --------
# MSM.eigenvectors_left
#
# """
# self._ensure_sample_eigendecomposition(k=k, ncv=ncv)
# return np.mean(self.sample_Ls, axis=0)
#
# def eigenvectors_left_std(self, k=None, ncv=None):
# """Sample standard deviation for the left eigenvectors.
#
# See also
# --------
# MSM.eigenvectors_left
#
# """
# self._ensure_sample_eigendecomposition(k=k, ncv=ncv)
# return np.std(self.sample_Ls, axis=0)
#
# def eigenvectors_left_conf(self, k=None, ncv=None):
# """Sample confidence interval for the left eigenvectors.
#
# See also
# --------
# MSM.eigenvectors_left
#
# """
# self._ensure_sample_eigendecomposition(k=k, ncv=ncv)
# return stat.confidence_interval(self.sample_Ls, alpha=self._confidence)
#
#
# # def eigenvectors_right_mean(self, k=None, ncv=None):
# # """Sample mean for the right eigenvectors.
# #
# # See also
# # --------
# # MSM.eigenvectors_right
# #
# # """
# # self._ensure_sample_eigendecomposition(k=k, ncv=ncv)
# # return np.mean(self.sample_Rs, axis=0)
# #
# # def eigenvectors_right_std(self, k=None, ncv=None):
# # """Sample standard deviation for the right eigenvectors.
# #
# # See also
# # --------
# # MSM.eigenvectors_right
# #
# # """
# # self._ensure_sample_eigendecomposition(k=k, ncv=ncv)
# # return np.std(self.sample_Rs, axis=0)
# #
# # def eigenvectors_right_conf(self, k=None, ncv=None):
# # """Sample confidence interval for the right eigenvectors.
# #
# # See also
# # --------
# # MSM.eigenvectors_right
# #
# # """
# # self._ensure_sample_eigendecomposition(k=k, ncv=ncv)
# # return stat.confidence_interval_arr(self.sample_Rs, alpha=self._confidence)
#
# def _sample_timescales(self):
# """Compute sample timescales from the sample eigenvalues"""
# res = np.empty((self._nsample), dtype=np.object)
# for i in range(self._nsample):
# res[i] = -self._lag / np.log(np.abs(self.sample_eigenvalues[i][1:]))
# return res
#
# def timescales_mean(self, k=None, ncv=None):
# """Sample mean for the timescales.
#
# See also
# --------
# MSM.timescales
#
# """
# self._ensure_sample_eigendecomposition(k=k, ncv=ncv)
# return np.mean(self._sample_timescales(), axis=0)
#
# def timescales_std(self, k=None, ncv=None):
# """Sample standard deviation for the timescales.
#
# See also
# --------
# MSM.timescales
#
# """
# self._ensure_sample_eigendecomposition(k=k, ncv=ncv)
# return np.std(self._sample_timescales(), axis=0)
#
# def timescales_conf(self, k=None, ncv=None):
# """Sample confidence interval for the timescales.
#
# See also
# --------
# MSM.timescales
#
# """
# self._ensure_sample_eigendecomposition(k=k, ncv=ncv)
# return stat.confidence_interval(self._sample_timescales(), alpha=self._confidence)
#
#
# def _sample_mfpt(self, A, B):
# """Compute sample timescales from the sample eigenvalues"""
# res = np.zeros((self._nsample))
# for i in range(self._nsample):
# res[i] = self._mfpt(self.sample_Ps[i], A, B, mu=self.sample_mus[i])
# return res
#
# def mfpt_mean(self, A, B):
# """Sample mean for the A->B mean first passage time.
#
# See also
# --------
# MSM.mfpt
#
# """
# return np.mean(self._sample_mfpt(A,B), axis=0)
#
# def mfpt_std(self, A, B):
# """Sample standard deviation for the A->B mean first passage time.
#
# See also
# --------
# MSM.mfpt
#
# """
# return np.std(self._sample_mfpt(A,B), axis=0)
#
# def mfpt_conf(self, A, B):
# """Sample confidence interval for the A->B mean first passage time.
#
# See also
# --------
# MSM.mfpt
#
# """
# return stat.confidence_interval(self._sample_mfpt(A,B), alpha=self._confidence)
#
# def _sample_committor_forward(self, A, B):
# """Compute sample timescales from the sample eigenvalues"""
# res = np.empty((self._nsample), dtype=np.object)
# for i in range(self._nsample):
# res[i] = self._committor_forward(self.sample_Ps[i], A, B)
# return res
#
# def committor_forward_mean(self, A, B):
# """Sample mean for the A->B forward committor.
#
# See also
# --------
# MSM.committor_forward
#
# """
# return np.mean(self._sample_committor_forward(A,B), axis=0)
#
# def committor_forward_std(self, A, B):
# """Sample standard deviation for the A->B forward committor.
#
# See also
# --------
# MSM.committor_forward
#
# """
# return np.std(self._sample_committor_forward(A,B), axis=0)
#
# def committor_forward_conf(self, A, B):
# """Sample confidence interval for the A->B forward committor.
#
# See also
# --------
# MSM.committor_forward
#
# """
# return stat.confidence_interval(self._sample_committor_forward(A,B), alpha=self._confidence)
#
#
# def _sample_committor_backward(self, A, B):
# """Compute sample timescales from the sample eigenvalues"""
# res = np.empty((self._nsample), dtype=np.object)
# for i in range(self._nsample):
# res[i] = self._committor_backward(self.sample_Ps[i], A, B, mu=self.sample_mus[i])
# return res
#
# def committor_backward_mean(self, A, B):
# """Sample mean for the A->B backward committor.
#
# See also
# --------
# MSM.committor_backward
#
# """
# return np.mean(self._sample_committor_backward(A,B), axis=0)
#
# def committor_backward_std(self, A, B):
# """Sample standard deviation for the A->B backward committor.
#
# See also
# --------
# MSM.committor_backward
#
# """
# return np.std(self._sample_committor_backward(A,B), axis=0)
#
# def committor_backward_conf(self, A, B):
# """Sample confidence interval for the A->B backward committor.
#
# See also
# --------
# MSM.committor_backward
#
# """
# return stat.confidence_interval(self._sample_committor_backward(A,B), alpha=self._confidence)
def __init__(self, dtrajs, dt_traj, lag, connectivity, active_set,
connected_sets, C_full, C_active, transition_matrix):
r"""Estimates a Markov model from discrete trajectories.
Parameters
----------
dtrajs : list containing ndarrays(dtype=int) or ndarray(n, dtype=int)
discrete trajectories, stored as integer ndarrays (arbitrary size)
or a single ndarray for only one trajectory.
dt_traj : str, optional, default='1 step'
Description of the physical time corresponding to the trajectory time
step. May be used by analysis algorithms such as plotting tools to
pretty-print the axes. By default '1 step', i.e. there is no physical
time unit. Specify by a number, whitespace and unit. Permitted units
are (* is an arbitrary string):
| 'fs', 'femtosecond*'
| 'ps', 'picosecond*'
| 'ns', 'nanosecond*'
| 'us', 'microsecond*'
| 'ms', 'millisecond*'
| 's', 'second*'
lagtime : int
lagtime for the MSM estimation in multiples of trajectory steps
connectivity : str, optional, default = 'largest'
Connectivity mode. Three methods are intended (currently only 'largest' is implemented)
* 'largest' : The active set is the largest reversibly connected set. All estimation will be done on this
subset and all quantities (transition matrix, stationary distribution, etc) are only defined on this
subset and are correspondingly smaller than the full set of states
* 'all' : The active set is the full set of states. Estimation will be conducted on each reversibly connected
set separately. That means the transition matrix will decompose into disconnected submatrices,
the stationary vector is only defined within subsets, etc. Currently not implemented.
* 'none' : The active set is the full set of states. Estimation will be conducted on the full set of states
without ensuring connectivity. This only permits nonreversible estimation. Currently not implemented.
active_set :
connected_sets :
C_full :
C_active :
transition_matrix :
"""
# superclass constructor
MSM.__init__(self,
transition_matrix,
dt_model=TimeUnit(dt_traj).get_scaled(lag))
# Making copies because we don't know what will happen to the arguments after this call
self.lag = lag
self.connectivity = copy.deepcopy(connectivity)
self.active_set = copy.deepcopy(active_set)
self._dtrajs_full = copy.deepcopy(dtrajs)
self.dt_traj = dt_traj
self._C_full = copy.deepcopy(C_full)
self._C_active = copy.deepcopy(C_active)
self._connected_sets = copy.deepcopy(connected_sets)
# calculate secondary quantities
self._nstates_full = np.shape(C_full)[0]
# full2active mapping
self._full2active = -1 * np.ones(self._nstates_full, dtype=int)
self._full2active[self._active_set] = np.array(list(
range(len(self._active_set))),
dtype=int)
# is estimated
self._is_estimated = True
