def __init__(self,
n,
m,
d=1,
A=None,
means=None,
covars=None,
w=None,
pi=None,
min_std=0.01,
init_type='uniform',
precision=numpy.double,
verbose=False):
'''
Construct a new Continuous HMM.
In order to initialize the model with custom parameters,
pass values for (A,means,covars,w,pi), and set the init_type to 'user'.
Normal initialization uses a uniform distribution for all probablities,
and is not recommended.
'''
_BaseHMM.__init__(self, n, m, precision, verbose) #@UndefinedVariable
self.d = d
self.A = A
self.pi = pi
self.means = means
self.covars = covars
self.w = w
self.min_std = min_std
self.reset(init_type=init_type)
def __init__(self,n,m,d=1,A=None,means=None,covars=None,w=None,pi=None,min_std=0.01,init_type='uniform',precision=numpy.double,verbose=False):
'''
Construct a new Continuous HMM.
In order to initialize the model with custom parameters,
pass values for (A,means,covars,w,pi), and set the init_type to 'user'.
Normal initialization uses a uniform distribution for all probablities,
and is not recommended.
'''
_BaseHMM.__init__(self,n,m,precision,verbose) #@UndefinedVariable
self.d = d
self.A = A
self.pi = pi
self.means = means
self.covars = covars
self.w = w
self.min_std = min_std
self.reset(init_type=init_type)
'''
flag to load some decoding info from cached files, for example bMap and durationLookup table .
makes decoding faster
'''
self.usePersistentFiles = True
self.logger = logging.getLogger(__name__)
self.logger.setLevel(logging.DEBUG)
def __init__(
self,
n,
m,
d=1,
A=None,
means=None,
covars=None,
w=None,
pi=None,
min_std=0.01,
init_type="uniform",
precision=numpy.double,
verbose=False,
):
"""
Construct a new Continuous HMM.
In order to initialize the model with custom parameters,
pass values for (A,means,covars,w,pi), and set the init_type to 'user'.
Normal initialization uses a uniform distribution for all probablities,
and is not recommended.
"""
_BaseHMM.__init__(self, n, m, precision, verbose) # @UndefinedVariable
self.d = d
self.A = A
self.pi = pi
self.means = means
self.covars = covars
self.w = w
self.min_std = min_std
self.reset(init_type=init_type)
def __init__(self,n,m,d=1,A=None,means=None,covars=None,w=None,pi=None,min_std=0.01,init_type='uniform',precision=numpy.double,verbose=False):
'''
Construct a new Continuous HMM.
In order to initialize the model with custom parameters,
pass values for (A,means,covars,w,pi), and set the init_type to 'user'.
Normal initialization uses a uniform distribution for all probablities,
and is not recommended.
'''
_BaseHMM.__init__(self,n,m,precision,verbose) #@UndefinedVariable
self.d = d
self.A = A
self.pi = pi
self.means = means
self.covars = covars
self.w = w
self.min_std = min_std
self.reset(init_type=init_type)
'''
flag to load some decoding info from cached files, for example bMap and durationLookup table
makes decoding faster
'''
self.usePersistentFiles = False
def __init__(self,
number_outputs,
n,
locations_per_segment,
start_t,
end_t,
number_latent_forces=1,
precision=np.double,
verbose=False):
assert n > 0
assert locations_per_segment > 0
assert number_outputs > 0
assert type(number_outputs) is type(1)
self.n = n # number of hidden states
self.number_outputs = number_outputs
self.start_t = start_t
self.end_t = end_t
self.sample_locations = np.linspace(start_t, end_t,
locations_per_segment)
self.locations_per_segment = locations_per_segment
# Pool of workers to perform parallel computations when need it.
self.pool = mp.Pool()
# covariance memoization
self.memo_covs = {}
self.number_latent_f = number_latent_forces
# initially not LFM params.
self.LFMparams = {}
# Latent Force Model objects
self.lfms = np.zeros(n, dtype='object')
for i in xrange(n):
self.lfms[i] = lfm2(self.number_latent_f, number_outputs)
self.lfms[i].set_inputs_with_same_ind(self.sample_locations)
self._INDEPENDENT_OUTPUTS = False
_BaseHMM.__init__(self, n, None, precision, verbose)
self.reset()
def __init__(self,
number_outputs,
n,
locations_per_segment,
start_t,
end_t,
precision=np.double,
verbose=False):
assert n > 0
assert locations_per_segment > 0
assert number_outputs > 0
assert type(number_outputs) is type(1)
self.n = n # number of hidden states
self.number_outputs = number_outputs
self.start_t = start_t
self.end_t = end_t
self.sample_locations = np.linspace(start_t, end_t,
locations_per_segment)
self.locations_per_segment = locations_per_segment
# Pool of workers to perform parallel computations when need it.
self.pool = mp.Pool()
# covariance memoization
self.memo_covs = {}
self.ICMparams = {}
# different icms.
self.icms = np.zeros(n, dtype='object')
for i in xrange(n):
self.icms[i] = icm(number_outputs, self.sample_locations)
_BaseHMM.__init__(self, n, None, precision, verbose)
self.reset()
def __init__(self,n,m,A=None,B=None,pi=None,init_type='uniform',precision=numpy.double,verbose=False):
'''
Construct a new Discrete HMM.
In order to initialize the model with custom parameters,
pass values for (A,B,pi), and set the init_type to 'user'.
Normal initialization uses a uniform distribution for all probablities,
and is not recommended.
'''
_BaseHMM.__init__(self,n,m,precision,verbose) #@UndefinedVariable
self.A = A
self.pi = pi
self.B = B
self.reset(init_type=init_type)
def __init__(self,
n,
m,
A=None,
B=None,
pi=None,
init_type='uniform',
precision=numpy.double,
verbose=False):
'''
Construct a new Discrete HMM.
In order to initialize the model with custom parameters,
pass values for (A,B,pi), and set the init_type to 'user'.
Normal initialization uses a uniform distribution for all probablities,
and is not recommended.
'''
_BaseHMM.__init__(self, n, m, precision, verbose) #@UndefinedVariable
self.A = A
self.pi = pi
self.B = B
self.reset(init_type=init_type)
