def test_dishonest_casino_larger_transition_p(self):
'''Dishonest Casino Example.'''
# Create transition probability matrix
# Create set of all observable symbols
V = [1, 2, 3, 4, 5, 6]
# Instantiate an HMM, note Pi is uniform probability distribution by default
m = hmmt.HMM(2, A=A2, B=B, V=V)
Obs = [ 1, 2, 3, 4, 5, 2, 1, 6, 6, 6, 5, 6 ]
log_prob_Obs, Alpha, c = hmmt.forward(m, Obs, scaling=1)
assert_almost_equal(log_prob_Obs, -20.12266, decimal=3, err_msg='Wrong observation probability')
Q_star, _, _ = hmmt.viterbi(m, Obs, scaling=1)
assert_equal(Q_star, [0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1], err_msg='Wrong Viterbi path')
Beta = hmmt.backward(m, Obs, c)
Gamma, Q_star = hmmt.individually_optimal_states(Alpha, Beta)
assert_almost_equal(Gamma,
[[0.8187172809335708, 0.8482909847967097, 0.8526704789882301, 0.8332628435868141, 0.7838346435680826, 0.6885099864086844, 0.5166714952004011, 0.21280359871744636, 0.11993462067436834, 0.10785580156479256, 0.15963081036800023, 0.1497144396249583],
[0.1812827190664292, 0.1517090152032903, 0.14732952101176994, 0.16673715641318582, 0.21616535643191742, 0.31149001359131556, 0.4833285047995989, 0.7871964012825536, 0.8800653793256317, 0.8921441984352074, 0.8403691896319997, 0.8502855603750418]],
decimal=5, err_msg='Wrong state probabilities')
assert_almost_equal(np.sum(Gamma[np.array([0,1]), :], axis=0), np.ones(len(Obs)),
decimal=3, err_msg='Gammas must sum up to 1 at each time')
assert_equal(Q_star, [0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1], 'Wrong individually-optimal states')
def __calculate_prob_viterbi(self):
'''Return the posterior IBD probability at each marker using the Viterbi path.
Here it is 1 if the state is an IBD>=1 state, otherwise 0. Sets the self.Q_star field.'''
# Remove observations that are IBS=0 - debugging
Q_star, _, _ = hmmt.viterbi(self.m, self.Obs, scaling=1)
self.Q_star = Q_star
self.p_ibd_viterbi = (Q_star == 0) | (Q_star == 2) | (Q_star == 4) | (Q_star == 6) | (Q_star == 7)
return self.p_ibd_viterbi
def __calculate_prob_viterbi(self):
'''Return the posterior IBD probability at each marker using the Viterbi path.
Here it is 1 if the state is an IBD>=1 state, otherwise 0. Sets the self.Q_star field.'''
# Remove observations that are IBS=0 - debugging
Q_star, _, _ = hmmt.viterbi(self.m, self.Obs, scaling=1)
self.Q_star = Q_star
self.p_ibd_viterbi = (Q_star == 1)
return self.p_ibd_viterbi
def test_dishonest_casino(self):
'''Dishonest Casino Example.'''
# Create set of all observable symbols
V = [1, 2, 3, 4, 5, 6]
# Instantiate an HMM, note Pi is uniform probability distribution by default
m = hmmt.HMM(2, A=A, B=B, V=V)
Obs = [ 1, 2, 3, 4, 5, 2, 1, 6, 6, 6, 5, 6 ]
log_prob_Obs, Alpha, c = hmmt.forward(m, Obs, scaling=1)
assert_almost_equal(log_prob_Obs, -20.918331960920177, decimal=5, err_msg='Wrong observation probability')
Q_star, _, _ = hmmt.viterbi(m, Obs, scaling=1)
assert_equal(Q_star, np.array([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]), 'Wrong Viterbi path')
Beta = hmmt.backward(m, Obs, c)
Gamma, Q_star = hmmt.individually_optimal_states(Alpha, Beta)
assert_almost_equal(Gamma,
[[0.6497764611558418, 0.6476464636785156, 0.6395341323290045, 0.6232731859090421, 0.594520672835795, 0.5455970624010458, 0.4634351125974832, 0.3259525777265338, 0.28060687540229046, 0.2668719322118299, 0.26641411220848843, 0.26052432611296034], [0.35022353884415813, 0.3523535363214843, 0.3604658676709956, 0.3767268140909578, 0.405479327164205, 0.4544029375989542, 0.5365648874025167, 0.6740474222734663, 0.7193931245977097, 0.7331280677881701, 0.7335858877915117, 0.7394756738870396]],
decimal=5, err_msg='Wrong state probabilities')
assert_equal(Q_star, [0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1], 'Wrong individually-optimal states')