def fromJSON(self, jsonObj):
"""Initializes the prototype with a JSON dictionary."""
super(PrototypeHMM, self).fromJSON(jsonObj)
self.N = jsonObj["N"]
self.model = WeightedGaussianHMM(self.N, "diag", algorithm="map", params="mc")
self.model.n_features = jsonObj["n_features"]
self.model.transmat_ = normalize(jsonObj["transmat"], axis=1)
self.model.startprob_ = normalize(jsonObj["startprob"], axis=0)
self.model._means_ = np.asarray(jsonObj["means"])
self.model._covars_ = np.asarray(jsonObj["covars"])
def train(self, obs, obs_weights=None, max_N=15):
"""Estimates the prototype from a set of observations.
Parameters
----------
max_N : int
The maximum lenght of the HMM.
"""
if obs_weights is None:
obs_weights = np.ones(len(obs))
else:
obs_weights = np.asarray(obs_weights)
# set the number of states
if self.num_states >= 1.0:
self.N = int(self.num_states)
else:
mean_length = np.mean([each_obs.shape[0] for each_obs in obs])
self.N = min(int(self.num_states * mean_length), max_N)
# transition prob: left-to-right
self.transmat = np.zeros((self.N, self.N))
for i in range(self.N):
self.transmat[i, i] = self.self_transprob
if i + 1 < self.N:
self.transmat[i, i + 1] = self.next_transprob
for j in range(i + 2, self.N):
self.transmat[i, j] = self.skip_transprob
self.transmat = normalize(self.transmat, axis=1)
# state prior prob: left-most only
self.startprob = np.zeros(self.N)
self.startprob[0] = 1.0
self.model = WeightedGaussianHMM(self.N, "diag", self.startprob, self.transmat, algorithm="map", params="mc")
self.num_obs = len(obs)
return self.model.fit(obs, obs_weights=obs_weights)
class PrototypeHMM(_Prototype):
"""HMM-based prototype.
This class uses HMM as the underlying model. The similarity is defined
in term of the log likelihood.
Parameters
----------
num_states : float
Number of hidden states. If num_states < 1, the number of states is
set proportionally to the average length of the observations.
For example, if num_states = 0.5, the number of states will
be set to 0.5 * average length of the observations.
self_transprob : float
Probability of staying in the same state.
next_transprob : float
Probability of moving to the adjacent state.
skip_transprob : float
Probability of moving to any other non-adjacent states.
Attributes
----------
N : int
Number of hidden states in the model.
"""
def __init__(self, label, num_states=0.5, self_transprob=0.8, next_transprob=0.2, skip_transprob=1e-6):
_Prototype.__init__(self, label)
self.num_states = num_states
self.self_transprob = self_transprob
self.next_transprob = next_transprob
self.skip_transprob = skip_transprob
def train(self, obs, obs_weights=None, max_N=15):
"""Estimates the prototype from a set of observations.
Parameters
----------
max_N : int
The maximum lenght of the HMM.
"""
if obs_weights is None:
obs_weights = np.ones(len(obs))
else:
obs_weights = np.asarray(obs_weights)
# set the number of states
if self.num_states >= 1.0:
self.N = int(self.num_states)
else:
mean_length = np.mean([each_obs.shape[0] for each_obs in obs])
self.N = min(int(self.num_states * mean_length), max_N)
# transition prob: left-to-right
self.transmat = np.zeros((self.N, self.N))
for i in range(self.N):
self.transmat[i, i] = self.self_transprob
if i + 1 < self.N:
self.transmat[i, i + 1] = self.next_transprob
for j in range(i + 2, self.N):
self.transmat[i, j] = self.skip_transprob
self.transmat = normalize(self.transmat, axis=1)
# state prior prob: left-most only
self.startprob = np.zeros(self.N)
self.startprob[0] = 1.0
self.model = WeightedGaussianHMM(self.N, "diag", self.startprob, self.transmat, algorithm="map", params="mc")
self.num_obs = len(obs)
return self.model.fit(obs, obs_weights=obs_weights)
def score(self, obs, last_state_only=True):
"""Calculates the score of an observation.
Returns
-------
score : float
If last_state_only=False, the score is defined as the log
likelihood of the observation under the model. Otherwise,
the score is defined as the log likelihood at the last state
only.
"""
obs = np.asarray(obs)
framelogprob = self.model._compute_log_likelihood(obs)
if last_state_only:
_, fwdlattice = self.model._do_forward_pass(framelogprob)
return fwdlattice[-1, -1]
else:
logprob, _ = self.model._do_forward_pass(framelogprob)
return logprob
def toJSON(self):
"""Returns a JSON dictionary representing the prototype."""
info = super(PrototypeHMM, self).toJSON()
info["n_components"] = int(self.model.n_components)
info["n_features"] = int(self.model.n_features)
info["transmat"] = self.model.transmat_.astype(np.float16).tolist()
info["startprob"] = self.model.startprob_.astype(np.float16).tolist()
info["means"] = self.model._means_.astype(np.float16).tolist()
info["covars"] = self.model._covars_.astype(np.float16).tolist()
info["N"] = self.N
return info
def fromJSON(self, jsonObj):
"""Initializes the prototype with a JSON dictionary."""
super(PrototypeHMM, self).fromJSON(jsonObj)
self.N = jsonObj["N"]
self.model = WeightedGaussianHMM(self.N, "diag", algorithm="map", params="mc")
self.model.n_features = jsonObj["n_features"]
self.model.transmat_ = normalize(jsonObj["transmat"], axis=1)
self.model.startprob_ = normalize(jsonObj["startprob"], axis=0)
self.model._means_ = np.asarray(jsonObj["means"])
self.model._covars_ = np.asarray(jsonObj["covars"])