def small_model():
n_match_states = 1
n_ins_states = 1
n_hstates = n_match_states + 2 * n_ins_states
n_simbols = 4
matchprob = .22
unmatchprob = .01
insertprob = .25
initprob_match = 0.6
initprob_ins = 0.2
initprob = np.array([initprob_match, initprob_ins, initprob_ins])
transprob = np.array([[.8, .1, .1], [.6, .4, 0], [.6, 0, .4]])
match_prob = np.ones((n_simbols, n_simbols)) * unmatchprob
for i in range(n_simbols):
match_prob[i, i] = matchprob
emitprob = np.ones((n_hstates, n_simbols, n_simbols)) * insertprob
emitprob[0, :, :] = match_prob
return PHMM(n_match_states=n_match_states,
n_xins_states=n_ins_states,
n_yins_states=n_ins_states,
initprob=initprob,
transprob=transprob,
emitprob=emitprob,
n_simbols=n_simbols)
def main():
"""
sample from small model and fit to PHMM
start from true props
"""
print(__name__)
print(inspect.getsource(main))
smodel = small_model()
fmodel = PHMM(n_match_states=1,
n_xins_states=1,
n_yins_states=1,
initprob=smodel._initprob,
transprob=smodel._transprob,
emitprob=smodel._emitprob)
xseqs, yseqs = sample_from_model(smodel, n_samples=200, len_seq=30)
fmodel.fit(xseqs, yseqs, max_iter=10000, verbose=True)
def large_model():
n_match_states = 3
n_ins_states = 2
n_hstates = n_match_states + 2 * n_ins_states
n_simbols = 4
initprob = np.array([.5, .1, .2, .08, .02, .08, .02])
transprob = np.array([[.58, .1, .1, .04, .02, .04, .12],
[.05, .9, .01, .01, .01, .01, .01],
[.1, .3, .1, .24, .1, .14, .02],
[.5, .1, .1, .1, 0, 0, 0], [.1, .3, .2, 0, .4, 0, 0],
[.1, .3, .5, 0, 0, .1, 0], [.2, .2, .2, 0, 0, 0,
.4]])
match1 = np.array([[.1, 0, 0, 0], [0, .1, 0, 0], [0, 0, .1, 0],
[0, 0, 0, .8]])
match2 = np.array([[.07, .01, .01, .01], [.01, .07, .01, .01],
[.01, .01, .57, .01], [.01, .01, .01, .17]])
match3 = np.array([[.37, .01, .01, .01], [.01, .37, .01, .01],
[.01, .01, .07, .01], [.01, .01, .01, .07]])
ins_a_x = np.ones(
(n_simbols, n_simbols)) * np.array([.5, .4, .05, .05])[:, np.newaxis]
ins_a_y = ins_a_x.T
ins_b_x = np.ones(
(n_simbols, n_simbols)) * np.array([.05, .05, .4, .5])[:, np.newaxis]
ins_b_y = ins_b_x.T
emitprob = np.zeros((n_hstates, n_simbols, n_simbols))
emitprob[0, :, :] = match1
emitprob[1, :, :] = match2
emitprob[2, :, :] = match3
emitprob[3, :, :] = ins_a_x
emitprob[4, :, :] = ins_b_x
emitprob[5, :, :] = ins_a_y
emitprob[6, :, :] = ins_b_y
return PHMM(n_match_states=n_match_states,
n_xins_states=n_ins_states,
n_yins_states=n_ins_states,
initprob=initprob,
transprob=transprob,
emitprob=emitprob,
n_simbols=n_simbols)
def med2_model():
n_match_states = 2
n_ins_states = 2
n_hstates = n_match_states + 2 * n_ins_states
n_simbols = 4
initprob = np.array([.3, .3, .05, .15, .05, .15])
transprob = np.array([[.78, .1, .04, .02, .04, .02],
[.1, .78, .04, .02, .04, .02],
[.45, .45, .1, 0, 0, 0], [.3, .3, 0, .4, 0, 0],
[.45, .45, 0, 0, .1, 0], [.3, .3, 0, 0, 0, .4]])
match1 = np.array([[.4, 0, 0, 0], [0, .4, 0, 0], [0, 0, .1, 0],
[0, 0, 0, .1]])
match2 = np.array([[.07, .01, .01, .01], [.01, .07, .01, .01],
[.01, .01, .37, .01], [.01, .01, .01, .37]])
ins_a_x = np.ones(
(n_simbols, n_simbols)) * np.array([.4, .3, .2, .1])[:, np.newaxis]
ins_a_y = ins_a_x.T
ins_b_x = np.ones(
(n_simbols, n_simbols)) * np.array([.1, .2, .3, .4])[:, np.newaxis]
ins_b_y = ins_b_x.T
emitprob = np.zeros((n_hstates, n_simbols, n_simbols))
emitprob[0, :, :] = match1
emitprob[1, :, :] = match2
emitprob[2, :, :] = ins_a_x
emitprob[3, :, :] = ins_b_x
emitprob[4, :, :] = ins_a_y
emitprob[5, :, :] = ins_b_y
return PHMM(n_match_states=n_match_states,
n_xins_states=n_ins_states,
n_yins_states=n_ins_states,
initprob=initprob,
transprob=transprob,
emitprob=emitprob,
n_simbols=n_simbols)
def med_model():
n_match_states = 1
n_ins_states = 2
n_hstates = n_match_states + 2 * n_ins_states
n_simbols = 4
initprob = np.array([.4, .15, .15, .15, .15])
matchprob = 0.235
unmatchprob = 0.05
transprob = np.array([[.8, .07, .03, .07, .03], [.9, .1, 0, 0, 0],
[.1, 0, .9, 0, 0], [.9, 0, 0, .1, 0],
[.1, 0, 0, 0, .9]])
ins_a_x = np.ones(
(n_simbols, n_simbols)) * np.array([.5, .4, .05, .05])[:, np.newaxis]
ins_a_y = ins_a_x.T
ins_b_x = np.ones(
(n_simbols, n_simbols)) * np.array([.05, .05, .4, .5])[:, np.newaxis]
ins_b_y = ins_b_x.T
match_prob = np.ones((n_simbols, n_simbols)) * unmatchprob
for i in range(n_simbols):
match_prob[i, i] = matchprob
emitprob = np.ones((n_hstates, n_simbols, n_simbols))
emitprob[0, :, :] = match_prob
emitprob[1, :, :] = ins_a_x
emitprob[2, :, :] = ins_b_x
emitprob[3, :, :] = ins_a_y
emitprob[4, :, :] = ins_b_y
return PHMM(n_match_states=n_match_states,
n_xins_states=n_ins_states,
n_yins_states=n_ins_states,
initprob=initprob,
transprob=transprob,
emitprob=emitprob,
n_simbols=n_simbols)
def emit_only_model():
n_match_states = 2
n_ins_states = 2
n_hstates = n_match_states + 2 * n_ins_states
n_simbols = 4
initprob = np.array([.4, .6, 0, 0, 0, 0])
transprob = np.array([[.8, .2, 0, 0, 0, 0], [.2, .8, 0, 0, 0, 0],
[.9, .1, 0, 0, 0, 0], [.9, .1, 0, 0, 0, 0],
[.9, .1, 0, 0, 0, 0], [.9, .1, 0, 0, 0, 0]])
matchprob = .16
unmatchprob = .03
insertprob = .25
match_prob = np.ones((n_simbols, n_simbols)) * unmatchprob
for i in range(n_simbols):
match_prob[i, i] = matchprob
emitprob = np.ones((n_hstates, n_simbols, n_simbols)) * insertprob
emitprob[0, :, :] = match_prob
matchprob = .10
unmatchprob = .05
match_prob = np.ones((n_simbols, n_simbols)) * unmatchprob
for i in range(n_simbols):
match_prob[i, i] = matchprob
emitprob[1, :, :] = match_prob
return PHMM(n_match_states=n_match_states,
n_xins_states=n_ins_states,
n_yins_states=n_ins_states,
initprob=initprob,
transprob=transprob,
emitprob=emitprob,
n_simbols=n_simbols)
def med3_model():
n_match_states = 2
n_ins_states = 1
n_hstates = n_match_states + 2 * n_ins_states
n_simbols = 4
initprob = np.array([
.4,
.4,
.1,
.1,
])
transprob = np.array([[.88, .1, .01, .01], [.1, .7, .15, .15][.01, .68,
.31, 0],
[.01, 0.68, 0, .31]])
match1 = np.array([[.4, 0, 0, 0], [0, .4, 0, 0], [0, 0, .1, 0],
[0, 0, 0, .1]])
match2 = np.array([[.11, .01, .01, .01], [.01, .11, .01, .01],
[.01, .01, .33, .01], [.01, .01, .01, .33]])
emitprob = np.zeros((n_hstates, n_simbols, n_simbols)) * .25
emitprob[0, :, :] = match1
emitprob[1, :, :] = match2
return PHMM(n_match_states=n_match_states,
n_xins_states=n_ins_states,
n_yins_states=n_ins_states,
initprob=initprob,
transprob=transprob,
emitprob=emitprob,
n_simbols=n_simbols)
def main():
"""
PHMM med
sample from med mode and fit on FABPHMM without shrinkage
"""
print(__name__)
print(inspect.getsource(main))
smodel = med3_model()
xseqs, yseqs = sample_from_model(smodel, n_samples=200, len_seq=50)
fab_small = FABPHMM(n_match_states=1,
n_xins_states=1,
n_yins_states=1,
shrink_threshold=1e-2,
stop_threshold=1e-4,
shrink=False)
fab_med = FABPHMM(n_match_states=1,
n_xins_states=2,
n_yins_states=2,
shrink_threshold=1e-2,
stop_threshold=1e-4,
shrink=False)
phmm_small = PHMM(n_match_states=1,
n_xins_states=1,
n_yins_states=1,
stop_threshold=1e-4)
phmm_med = PHMM(n_match_states=1,
n_xins_states=2,
n_yins_states=2,
stop_threshold=1e-4)
fic_small = fab_small.fit(xseqs, yseqs, max_iter=500, verbose=True, verbose_level=1)
ll_small = fab_small.score(xseqs, yseqs, type="ll")
fic_med = fab_med.fit(xseqs, yseqs, max_iter=500, verbose=True, verbose_level=1)
ll_med = fab_med.score(xseqs, yseqs, type="ll")
phmm_small.fit(xseqs, yseqs, max_iter=500, verbose=True, verbose_level=1)
ll_phmms = phmm_small.score(xseqs, yseqs)
phmm_med.fit(xseqs, yseqs, max_iter=500, verbose=True, verbose_level=1)
ll_phmmm = phmm_med.score(xseqs, yseqs)
ll_org = smodel.score(xseqs, yseqs)
smodel.fit(xseqs, yseqs)
ll_org_fit = smodel.score(xseqs, yseqs)
print("dump fab_small")
fab_small._print_states(verbose_level=2)
print("dump fab_med")
fab_med._print_states(verbose_level=2)
print("dump phmm_small")
phmm_small._print_states(verbose_level=2)
print("dump phmm_med")
phmm_med._print_states(verbose_level=2)
print("fic_small", fic_small)
print("ll_small", ll_small)
print("fic_med", fic_med)
print("ll_med", ll_med)
print("ll_phmms", ll_phmms)
print("ll_phmmm", ll_phmmm)
print("ll_org", ll_org)
print("ll_org", ll_org_fit)
def m533_model():
n_match_states = 5
n_xins_states = 3
n_yins_states = 3
n_simbols = 4
initprob = np.array([
1.74025574e-001, 2.02487037e-001, 9.02697286e-002, 1.65972599e-001,
2.86802745e-001, 1.50094058e-002, 2.51820392e-002, 4.02508702e-002,
1.53121845e-039, 6.18009805e-135, 4.68295749e-071
])
transprob = np.array(
[[
2.29870373e-01, 2.36430363e-01, 8.40144731e-02, 1.73321989e-01,
1.96457664e-01, 5.96742117e-03, 3.33349159e-02, 8.15858501e-03,
1.87202145e-02, 8.59183666e-03, 5.13216415e-03
],
[
5.01312337e-12, 2.28952298e-01, 4.07828833e-01, 2.55801116e-01,
9.79197540e-02, 2.41478407e-03, 6.52019556e-07, 9.02461283e-09,
1.55197808e-03, 4.92408350e-03, 6.06492092e-04
],
[
3.07117325e-01, 2.35764445e-01, 4.43833037e-03, 1.23780006e-01,
1.87954692e-01, 7.78406044e-04, 6.88754357e-02, 1.14351806e-02,
4.52511180e-02, 1.46050616e-02, 3.21832829e-15
],
[
7.14116703e-01, 1.53998640e-01, 1.60955988e-04, 6.97953684e-08,
1.25508826e-01, 1.80708586e-07, 3.71067294e-04, 9.28372641e-04,
4.88889236e-03, 2.62923810e-05, 5.02089294e-10
],
[
5.17057140e-05, 1.66285304e-01, 3.13551009e-07, 1.95758241e-01,
5.84834189e-01, 9.88169375e-07, 1.99206013e-02, 1.33119134e-02,
6.18591453e-05, 1.92128225e-02, 5.62062197e-04
],
[
1.79693070e-02, 2.00950692e-02, 1.40127671e-20, 9.43293510e-03,
1.19826799e-06, 9.52501490e-01, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00
],
[
2.98527289e-02, 1.50393055e-01, 5.06471644e-03, 1.58615219e-01,
1.10361795e-01, 0.00000000e+00, 5.45712485e-01, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00
],
[
7.59442172e-03, 2.18456041e-02, 5.46901923e-03, 1.76449076e-02,
3.96863886e-02, 0.00000000e+00, 0.00000000e+00, 9.07759659e-01,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00
],
[
4.67442572e-02, 8.98702895e-02, 1.65756948e-06, 8.31517001e-02,
2.08608060e-11, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
7.80232096e-01, 0.00000000e+00, 0.00000000e+00
],
[
7.15712398e-03, 6.52943502e-02, 1.88623622e-07, 3.45519618e-02,
1.51034432e-01, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 7.41961943e-01, 0.00000000e+00
],
[
1.31535218e-02, 4.08240365e-03, 1.21362185e-08, 8.53321007e-04,
1.24688602e-02, 0.00000000e+00, 0.00000000e+00, 0.00000000e+00,
0.00000000e+00, 0.00000000e+00, 9.69441881e-01
]])
emitprob = np.array(
[[[1.12934790e-06, 5.11357725e-08, 1.35402425e-11, 5.09960520e-02],
[7.84216424e-03, 4.05916075e-04, 8.32608389e-07, 5.58257158e-02],
[2.09182444e-04, 9.68168139e-05, 3.35980195e-03, 1.58042086e-01],
[2.56129734e-02, 4.03960988e-02, 1.05814116e-01, 5.51397063e-01]],
[[3.39180663e-01, 1.03591025e-01, 5.55270296e-02, 1.03365937e-02],
[3.39376049e-02, 3.17637465e-05, 2.50436596e-03, 7.77639792e-24],
[4.04508063e-02, 1.00965494e-02, 3.37809949e-01, 9.65488908e-05],
[2.21679659e-02, 2.61446333e-03, 3.76767280e-02, 3.97794312e-03]],
[[1.92812813e-02, 4.19521933e-02, 2.76722186e-02, 1.15564816e-04],
[7.23901117e-02, 1.13845299e-03, 3.93167074e-02, 3.49112068e-25],
[3.57959261e-02, 5.94701879e-02, 6.47307074e-01, 5.55602258e-02],
[3.52450725e-11, 3.43522138e-18, 5.58084883e-08, 1.40277494e-14]],
[[1.34720910e-19, 6.57493920e-02, 1.23460351e-02, 6.04621213e-04],
[3.41879316e-13, 4.19824798e-01, 4.06314768e-02, 1.06237294e-02],
[3.34247036e-21, 3.87406507e-02, 2.82440687e-01, 1.06737312e-06],
[2.82413472e-04, 2.97719200e-02, 8.21702019e-02, 1.68130066e-02]],
[[2.06528696e-01, 6.20537625e-02, 1.27373567e-02, 2.24556009e-03],
[8.65655521e-02, 4.22946647e-01, 2.16098830e-02, 9.36881641e-03],
[1.34106238e-02, 1.24677709e-02, 8.08148578e-02, 4.77177240e-03],
[1.21511972e-02, 1.28660242e-02, 3.22060210e-02, 7.25545857e-03]],
[[3.53924437e-01, 3.53924437e-01, 3.53924437e-01, 3.53924437e-01],
[1.36852092e-01, 1.36852092e-01, 1.36852092e-01, 1.36852092e-01],
[1.52331030e-01, 1.52331030e-01, 1.52331030e-01, 1.52331030e-01],
[3.56892441e-01, 3.56892441e-01, 3.56892441e-01, 3.56892441e-01]],
[[1.91320662e-01, 1.91320662e-01, 1.91320662e-01, 1.91320662e-01],
[2.23912507e-01, 2.23912507e-01, 2.23912507e-01, 2.23912507e-01],
[3.14486347e-01, 3.14486347e-01, 3.14486347e-01, 3.14486347e-01],
[2.70280484e-01, 2.70280484e-01, 2.70280484e-01, 2.70280484e-01]],
[[1.78773734e-01, 1.78773734e-01, 1.78773734e-01, 1.78773734e-01],
[3.46984152e-01, 3.46984152e-01, 3.46984152e-01, 3.46984152e-01],
[3.30984949e-01, 3.30984949e-01, 3.30984949e-01, 3.30984949e-01],
[1.43257165e-01, 1.43257165e-01, 1.43257165e-01, 1.43257165e-01]],
[[1.91986162e-01, 1.40829112e-01, 5.50749331e-01, 1.16435395e-01],
[1.91986162e-01, 1.40829112e-01, 5.50749331e-01, 1.16435395e-01],
[1.91986162e-01, 1.40829112e-01, 5.50749331e-01, 1.16435395e-01],
[1.91986162e-01, 1.40829112e-01, 5.50749331e-01, 1.16435395e-01]],
[[1.62355215e-01, 5.29650568e-01, 1.69767030e-01, 1.38227187e-01],
[1.62355215e-01, 5.29650568e-01, 1.69767030e-01, 1.38227187e-01],
[1.62355215e-01, 5.29650568e-01, 1.69767030e-01, 1.38227187e-01],
[1.62355215e-01, 5.29650568e-01, 1.69767030e-01, 1.38227187e-01]],
[[2.55109862e-01, 2.07452002e-01, 2.16013242e-01, 3.21424894e-01],
[2.55109862e-01, 2.07452002e-01, 2.16013242e-01, 3.21424894e-01],
[2.55109862e-01, 2.07452002e-01, 2.16013242e-01, 3.21424894e-01],
[2.55109862e-01, 2.07452002e-01, 2.16013242e-01, 3.21424894e-01]]])
return PHMM(n_match_states=n_match_states,
n_xins_states=n_xins_states,
n_yins_states=n_yins_states,
initprob=initprob,
transprob=transprob,
emitprob=emitprob,
n_simbols=n_simbols)
def huge_model():
n_match_states = 5
n_xins_states = 2
n_yins_states = 1
n_simbols = 4
initprob = np.array([
1.52250209e-01, 1.73901308e-01, 2.12584427e-01, 2.23262778e-01,
1.51633394e-01, 4.38939224e-02, 4.24739609e-02, 2.13068007e-26
])
transprob = np.array(
[[
8.10682058e-02, 5.20547306e-01, 2.77782021e-01, 4.79388383e-05,
5.73136913e-02, 1.91572066e-02, 2.28284185e-02, 2.12552123e-02
],
[
3.23034774e-01, 2.90914732e-01, 2.99734944e-01, 1.31605054e-02,
2.14674649e-02, 2.55636094e-02, 1.08771769e-02, 1.52467930e-02
],
[
6.29750882e-01, 5.99684915e-08, 3.21309829e-01, 1.65327737e-04,
1.89662317e-02, 4.41917044e-03, 7.40436483e-03, 1.79841351e-02
],
[
2.52164440e-03, 4.66846900e-03, 1.00893950e-02, 8.65158118e-01,
7.53603488e-02, 2.98792557e-02, 6.25705646e-04, 1.16970631e-02
],
[
4.48524703e-02, 2.03525802e-02, 8.45472416e-02, 8.37729739e-02,
7.35089193e-01, 1.56998528e-02, 1.92334615e-03, 1.37623425e-02
],
[
7.88854838e-06, 1.04723049e-01, 4.88676478e-02, 8.48254908e-02,
1.02709792e-01, 6.58866132e-01, 0.00000000e+00, 0.00000000e+00
],
[
2.28500256e-03, 2.50986396e-02, 7.20435122e-02, 6.95896615e-05,
1.40030880e-03, 0.00000000e+00, 8.99102947e-01, 0.00000000e+00
],
[
6.85894986e-02, 2.80908651e-02, 6.05330847e-02, 3.21919966e-02,
3.27113854e-02, 0.00000000e+00, 0.00000000e+00, 7.77883170e-01
]])
emitprob = np.array(
[[[4.40086542e-03, 7.14642286e-02, 2.28578318e-02, 1.63966800e-03],
[1.70746562e-02, 3.88626536e-01, 4.55716187e-02, 5.27847677e-10],
[7.20770705e-03, 3.01580599e-02, 3.09579378e-01, 1.24607445e-02],
[3.80436237e-21, 2.44972642e-02, 6.08879067e-02, 3.57353442e-03]],
[[5.06862412e-01, 1.09158647e-01, 4.48658641e-02, 2.39580174e-02],
[1.36818034e-01, 5.35084894e-05, 7.20991883e-05, 1.33164573e-03],
[4.43529434e-02, 7.38204315e-03, 4.65462409e-02, 3.05850450e-04],
[4.52169843e-02, 4.02936706e-03, 2.84035704e-02, 6.42772321e-04]],
[[3.13647232e-06, 3.01377254e-02, 1.23360820e-04, 3.08615105e-02],
[3.00667529e-04, 1.76152900e-01, 7.29051762e-03, 4.30562956e-02],
[7.34084625e-06, 3.24690923e-05, 1.48346629e-02, 1.09601252e-01],
[1.69782347e-02, 5.34127994e-02, 9.55213896e-02, 4.21685737e-01]],
[[9.07112819e-02, 3.83118861e-02, 1.77610784e-02, 4.71015547e-03],
[5.94417277e-02, 4.15133866e-01, 3.30740455e-02, 2.94775779e-02],
[1.68330679e-02, 2.29128233e-02, 1.60927935e-01, 2.76283268e-02],
[3.18583945e-03, 1.49696712e-02, 3.11308911e-02, 3.37898255e-02]],
[[1.65057479e-02, 1.60187447e-02, 1.54342918e-02, 1.19093828e-02],
[2.37819064e-02, 5.81066977e-02, 2.63616219e-02, 1.13652863e-02],
[2.87327021e-02, 3.53167411e-02, 5.00302181e-01, 7.36873815e-02],
[1.15195462e-02, 9.07015696e-03, 6.61471172e-02, 9.57404942e-02]],
[[1.72774419e-01, 1.72774419e-01, 1.72774419e-01, 1.72774419e-01],
[3.68970649e-01, 3.68970649e-01, 3.68970649e-01, 3.68970649e-01],
[3.52549866e-01, 3.52549866e-01, 3.52549866e-01, 3.52549866e-01],
[1.05705066e-01, 1.05705066e-01, 1.05705066e-01, 1.05705066e-01]],
[[2.43306155e-01, 2.43306155e-01, 2.43306155e-01, 2.43306155e-01],
[2.31434163e-01, 2.31434163e-01, 2.31434163e-01, 2.31434163e-01],
[2.62364505e-01, 2.62364505e-01, 2.62364505e-01, 2.62364505e-01],
[2.62895176e-01, 2.62895176e-01, 2.62895176e-01, 2.62895176e-01]],
[[1.96393214e-01, 2.91945036e-01, 3.04948763e-01, 2.06712987e-01],
[1.96393214e-01, 2.91945036e-01, 3.04948763e-01, 2.06712987e-01],
[1.96393214e-01, 2.91945036e-01, 3.04948763e-01, 2.06712987e-01],
[1.96393214e-01, 2.91945036e-01, 3.04948763e-01, 2.06712987e-01]]])
return PHMM(n_match_states=n_match_states,
n_xins_states=n_xins_states,
n_yins_states=n_yins_states,
initprob=initprob,
transprob=transprob,
emitprob=emitprob,
n_simbols=n_simbols)