def set_global_params(self,
K=0,
beta=None,
U1=None,
U0=None,
Ebeta=None,
EbetaLeftover=None,
**kwargs):
''' Directly set global parameter vector v
using provided arguments
'''
self.K = K
if U1 is not None and U0 is not None:
self.v = U1 / (U1 + U0)
if Ebeta is not None and EbetaLeftover is not None:
Ebeta = np.squeeze(Ebeta)
EbetaLeftover = np.squeeze(EbetaLeftover)
beta = np.hstack([Ebeta, EbetaLeftover])
elif beta is not None:
assert beta.size == K
beta = np.hstack([beta, np.min(beta) / 100.])
beta = beta / np.sum(beta)
if beta is not None:
assert abs(np.sum(beta) - 1.0) < 0.005
self.v = OptimPE.beta2v(beta)
assert self.v.size == self.K
self.set_helper_params()
def update_global_params_VB(self, SS, **kwargs):
''' Update global parameters v that control topic probabilities beta
'''
self.K = SS.K
sumLogPi = np.hstack([SS.sumLogPiActive, SS.sumLogPiUnused])
assert sumLogPi.size == self.K + 1
try:
v, fofv, Info = OptimPE.estimate_v(alpha0=self.alpha0, gamma=self.gamma,
sumLogPi=sumLogPi, nDoc=SS.nDoc)
self.v = v
except ValueError as v:
if str(v).count('failed') > 0:
if self.v.size != self.K:
beta = np.hstack([SS.N, 0.01])
beta /= beta.sum()
self.v = OptimPE.beta2v(beta)
else:
pass # keep current estimate of v!
assert self.v.size == self.K
self.set_helper_params()
def update_global_params_VB(self, SS, **kwargs):
''' Update global parameters v that control topic probabilities beta
'''
self.K = SS.K
sumLogPi = np.hstack([SS.sumLogPiActive, SS.sumLogPiUnused])
assert sumLogPi.size == self.K + 1
try:
v, fofv, Info = OptimPE.estimate_v(alpha0=self.alpha0,
gamma=self.gamma,
sumLogPi=sumLogPi,
nDoc=SS.nDoc)
self.v = v
except ValueError as v:
if str(v).count('failed') > 0:
if self.v.size != self.K:
beta = np.hstack([SS.N, 0.01])
beta /= beta.sum()
self.v = OptimPE.beta2v(beta)
else:
pass # keep current estimate of v!
assert self.v.size == self.K
self.set_helper_params()
def set_global_params(self, K=0, beta=None, U1=None, U0=None,
Ebeta=None, EbetaLeftover=None, **kwargs):
''' Directly set global parameter vector v
using provided arguments
'''
self.K = K
if U1 is not None and U0 is not None:
self.v = U1 / (U1 + U0)
if Ebeta is not None and EbetaLeftover is not None:
Ebeta = np.squeeze(Ebeta)
EbetaLeftover = np.squeeze(EbetaLeftover)
beta = np.hstack( [Ebeta, EbetaLeftover])
elif beta is not None:
assert beta.size == K
beta = np.hstack([beta, np.min(beta)/100.])
beta = beta/np.sum(beta)
if beta is not None:
assert abs(np.sum(beta) - 1.0) < 0.005
self.v = OptimPE.beta2v(beta)
assert self.v.size == self.K
self.set_helper_params()
def set_helper_params(self):
self.Ebeta = OptimPE.v2beta(self.v)
def set_helper_params(self): self.Ebeta = OptimPE.v2beta(self.v)
