def test_hssmm(self):
import numpy as np
from matplotlib import pyplot as plt
from pyhsmm.models import HSMMIntNegBinVariant
from pyhsmm.basic.models import MixtureDistribution
from pyhsmm.basic.distributions import Gaussian, NegativeBinomialIntegerRVariantDuration
from pyhsmm.util.text import progprint_xrange
#############################
# generate synthetic data #
#############################
states_in_hsmm = 5
components_per_GMM = 3
component_hyperparameters = dict(mu_0=np.zeros(2), sigma_0=np.eye(2), kappa_0=0.01, nu_0=3)
GMMs = [MixtureDistribution(
alpha_0=4.,
components=[Gaussian(**component_hyperparameters) for i in range(components_per_GMM)])
for state in range(states_in_hsmm)]
true_dur_distns = [
NegativeBinomialIntegerRVariantDuration(np.r_[0., 0, 0, 0, 0, 1, 1, 1], alpha_0=5., beta_0=5.)
for state in range(states_in_hsmm)]
truemodel = HSMMIntNegBinVariant(
init_state_concentration=10.,
alpha=6., gamma=6.,
obs_distns=GMMs,
dur_distns=true_dur_distns)
training_datas = [truemodel.generate(1000)[0] for i in range(5)]
test_data = truemodel.generate(5000)[0]
#####################################
# set up FrozenMixture components #
#####################################
# list of all Gaussians
component_library = [c for m in GMMs for c in m.components]
library_size = len(component_library)
# initialize weights to indicator on one component
init_weights = np.eye(library_size)
#obs_distns = [FrozenMixtureDistribution(
# components=component_library,
# alpha_0=4,
# weights=row)
# for row in init_weights]
################
# build HSMM #
################
dur_distns = [NegativeBinomialIntegerRVariantDuration(np.r_[0., 0, 0, 0, 0, 1, 1, 1], alpha_0=5., beta_0=5.)
for state in range(library_size)]
Gaussian(**component_hyperparameters)
for i in range(components_per_GMM)
]) for state in range(states_in_hsmm)
]
true_dur_distns = [
NegativeBinomialIntegerRVariantDuration(np.r_[0., 0, 0, 0, 0, 0, 1, 1, 1,
1],
alpha_0=5.,
beta_0=5.)
for state in range(states_in_hsmm)
]
truemodel = HSMMIntNegBinVariant(init_state_concentration=10.,
alpha=6.,
gamma=2.,
obs_distns=GMMs,
dur_distns=true_dur_distns)
training_datas = [truemodel.rvs(1000) for i in range(num_training_seqs)]
test_data = truemodel.rvs(1000)
#####################################
# set up FrozenMixture components #
#####################################
# list of all Gaussians
component_library = [c for m in GMMs for c in m.components]
library_size = len(component_library)
# initialize weights to indicator on one component
states_in_hsmm = 5
components_per_GMM = 3
component_hyperparameters = dict(mu_0=np.zeros(2),sigma_0=np.eye(2),kappa_0=0.01,nu_0=3)
GMMs = [MixtureDistribution(
alpha_0=4.,
components=[Gaussian(**component_hyperparameters) for i in range(components_per_GMM)])
for state in range(states_in_hsmm)]
true_dur_distns = [NegativeBinomialIntegerRVariantDuration(np.r_[0.,0,0,0,0,1,1,1],alpha_0=5.,beta_0=5.)
for state in range(states_in_hsmm)]
truemodel = HSMMIntNegBinVariant(
init_state_concentration=10.,
alpha=6.,gamma=6.,
obs_distns=GMMs,
dur_distns=true_dur_distns)
training_datas = [truemodel.generate(1000)[0] for i in range(5)]
test_data = truemodel.generate(5000)[0]
#####################################
# set up FrozenMixture components #
#####################################
# list of all Gaussians
component_library = [c for m in GMMs for c in m.components]
library_size = len(component_library)
# initialize weights to indicator on one component
components=[
Gaussian(**component_hyperparameters)
for i in range(components_per_GMM)
]) for state in range(states_in_hsmm)
]
true_dur_distns = [
NegativeBinomialIntegerRVariantDuration(np.r_[0., 0, 0, 1, 1, 1, 1, 1],
alpha_0=5.,
beta_0=5.)
for state in range(states_in_hsmm)
]
truemodel = HSMMIntNegBinVariant(init_state_concentration=10.,
alpha=6.,
gamma=6.,
obs_distns=GMMs,
dur_distns=true_dur_distns)
data, truelabels = truemodel.generate(10000)
#####################################
# set up FrozenMixture components #
#####################################
# list of all Gaussians
component_library = [c for m in GMMs for c in m.components]
library_size = len(component_library)
# initialize weights to indicator on one component
init_weights = np.eye(library_size)
states_in_hsmm = 5
components_per_GMM = 3
component_hyperparameters = dict(mu_0=np.zeros(2),sigma_0=np.eye(2),kappa_0=0.025,nu_0=3)
GMMs = [MixtureDistribution(
alpha_0=4.,
components=[Gaussian(**component_hyperparameters) for i in range(components_per_GMM)])
for state in range(states_in_hsmm)]
true_dur_distns = [NegativeBinomialIntegerRVariantDuration(np.r_[0.,0,0,1,1,1,1,1],alpha_0=5.,beta_0=5.)
for state in range(states_in_hsmm)]
truemodel = HSMMIntNegBinVariant(
init_state_concentration=10.,
alpha=6.,gamma=6.,
obs_distns=GMMs,
dur_distns=true_dur_distns)
data, truelabels = truemodel.generate(1000)
#################
# remove data #
#################
# using an HMM of course
deletion_hmm = HMM(
init_state_concentration=10.,
alpha=6.,gamma=6.,
obs_distns=GMMs[:2], # placeholders, not really used
components=[
Gaussian(**component_hyperparameters)
for i in range(components_per_GMM)
]) for state in range(states_in_hsmm)
]
true_dur_distns = [
NegativeBinomialIntegerRVariantDuration(np.r_[0., 0, 0, 1, 1, 1, 1, 1],
alpha_0=5.,
beta_0=5.)
for state in range(states_in_hsmm)
]
truemodel = HSMMIntNegBinVariant(init_state_concentration=10.,
alpha=6.,
gamma=6.,
obs_distns=GMMs,
dur_distns=true_dur_distns)
datas = [truemodel.generate(40000)[0][5:] for i in range(6)]
#####################################
# set up FrozenMixture components #
#####################################
# list of all Gaussians
component_library = [c for m in GMMs for c in m.components]
library_size = len(component_library)
Nmax = 4 * library_size
states_in_hsmm = 5
components_per_GMM = 3
component_hyperparameters = dict(mu_0=np.zeros(2),sigma_0=np.eye(2),kappa_0=0.01,nu_0=3)
GMMs = [MixtureDistribution(
alpha_0=4.,
components=[Gaussian(**component_hyperparameters) for i in range(components_per_GMM)])
for state in range(states_in_hsmm)]
true_dur_distns = [NegativeBinomialIntegerRVariantDuration(np.r_[0.,0,0,0,0,1,1,1],alpha_0=5.,beta_0=5.)
for state in range(states_in_hsmm)]
truemodel = HSMMIntNegBinVariant(
init_state_concentration=10.,
alpha=6.,gamma=6.,
obs_distns=GMMs,
dur_distns=true_dur_distns)
training_datas = {}
for group_id in range(2):
for i in range(2):
training_datas[group_id] = truemodel.generate(500)[0]
truemodel.trans_distn.resample() # new transition matrix for the next group
#####################################
# set up FrozenMixture components #
#####################################
# list of all Gaussians
component_library = [c for m in GMMs for c in m.components]
components=[
Gaussian(**component_hyperparameters)
for i in range(components_per_GMM)
]) for state in range(states_in_hsmm)
]
true_dur_distns = [
NegativeBinomialIntegerRVariantDuration(np.r_[0., 0, 0, 0, 0, 1, 1, 1],
alpha_0=5.,
beta_0=5.)
for state in range(states_in_hsmm)
]
truemodel = HSMMIntNegBinVariant(init_state_concentration=10.,
alpha=6.,
gamma=6.,
obs_distns=GMMs,
dur_distns=true_dur_distns)
training_datas = [truemodel.generate(1000)[0] for i in range(5)]
test_data = truemodel.generate(5000)[0]
#####################################
# set up FrozenMixture components #
#####################################
# list of all Gaussians
component_library = [c for m in GMMs for c in m.components]
library_size = len(component_library)
# initialize weights to indicator on one component
states_in_hsmm = 5
components_per_GMM = 3
component_hyperparameters = dict(mu_0=np.zeros(2),sigma_0=np.eye(2),kappa_0=0.025,nu_0=3)
GMMs = [MixtureDistribution(
alpha_0=4.,
components=[Gaussian(**component_hyperparameters) for i in range(components_per_GMM)])
for state in range(states_in_hsmm)]
true_dur_distns = [NegativeBinomialIntegerRVariantDuration(np.r_[0.,0,0,1,1,1,1,1],alpha_0=5.,beta_0=5.)
for state in range(states_in_hsmm)]
truemodel = HSMMIntNegBinVariant(
init_state_concentration=10.,
alpha=6.,gamma=6.,
obs_distns=GMMs,
dur_distns=true_dur_distns)
datas = [truemodel.generate(40000)[0][5:] for i in range(6)]
#####################################
# set up FrozenMixture components #
#####################################
# list of all Gaussians
component_library = [c for m in GMMs for c in m.components]
library_size = len(component_library)
Nmax = 4*library_size
states_in_hsmm = 5
components_per_GMM = 3
component_hyperparameters = dict(mu_0=np.zeros(2),sigma_0=np.eye(2),kappa_0=0.025,nu_0=3)
GMMs = [MixtureDistribution(
alpha_0=4.,
components=[Gaussian(**component_hyperparameters) for i in range(components_per_GMM)])
for state in range(states_in_hsmm)]
true_dur_distns = [NegativeBinomialIntegerRVariantDuration(np.r_[0.,0,0,0,0,0,1,1,1,1],alpha_0=5.,beta_0=5.)
for state in range(states_in_hsmm)]
truemodel = HSMMIntNegBinVariant(
init_state_concentration=10.,
alpha=6.,gamma=2.,
obs_distns=GMMs,
dur_distns=true_dur_distns)
training_datas = [truemodel.rvs(1000) for i in range(num_training_seqs)]
test_data = truemodel.rvs(1000)
#####################################
# set up FrozenMixture components #
#####################################
# list of all Gaussians
component_library = [c for m in GMMs for c in m.components]
library_size = len(component_library)
# initialize weights to indicator on one component
components=[
Gaussian(**component_hyperparameters)
for i in range(components_per_GMM)
]) for state in range(states_in_hsmm)
]
true_dur_distns = [
NegativeBinomialIntegerRVariantDuration(np.r_[0., 0, 0, 0, 0, 1, 1, 1],
alpha_0=5.,
beta_0=5.)
for state in range(states_in_hsmm)
]
truemodel = HSMMIntNegBinVariant(init_state_concentration=10.,
alpha=6.,
gamma=6.,
obs_distns=GMMs,
dur_distns=true_dur_distns)
training_datas = {}
for group_id in range(2):
for i in range(2):
training_datas[group_id] = truemodel.generate(500)[0]
truemodel.trans_distn.resample(
) # new transition matrix for the next group
#####################################
# set up FrozenMixture components #
#####################################
# list of all Gaussians