def step(self, anneal, model_params, my_data):
""" Perform an EM-step """
# Noisify model parameters
model_params = self.noisify_params(model_params, anneal)
# Sanity check model parameters
model_params = self.check_params(model_params)
# For partial EM-step: select batch
my_pdata = self.select_partial_data(anneal, my_data)
# Annotate partial dataset with hidden-state candidates
my_pdata = self.select_Hprimes(model_params, my_pdata)
# Do E-step and calculate joint-probabilities
my_joint_prob = self.E_step(anneal, model_params, my_pdata)
# Use joint-probabilities to derive new parameter set
new_model_params = self.M_step(anneal, model_params, my_joint_prob, my_pdata)
# Calculate Objecive Function
#[Q, theo_Q] = self.objective(my_joint_prob, model_params, my_pdata)
#dlog.append('Q', Q)
#dlog.append('theo_Q', theo_Q)
# Log iboth model parameters and annealing parameters
dlog.append_all(new_model_params)
dlog.append_all(anneal.as_dict())
return new_model_params
def step(self, anneal, model_params, my_data):
""" Perform an EM-step """
# Noisify model parameters
model_params = self.noisify_params(model_params, anneal)
# Sanity check model parameters
model_params = self.check_params(model_params)
# For partial EM-step: select batch
my_pdata = self.select_partial_data(anneal, my_data)
# Annotate partial dataset with hidden-state candidates
my_pdata = self.select_Hprimes(model_params, my_pdata)
# Do E-step and calculate joint-probabilities
my_joint_prob = self.E_step(anneal, model_params, my_pdata)
# Use joint-probabilities to derive new parameter set
new_model_params = self.M_step(anneal, model_params, my_joint_prob,
my_pdata)
# Calculate Objecive Function
#[Q, theo_Q] = self.objective(my_joint_prob, model_params, my_pdata)
#dlog.append('Q', Q)
#dlog.append('theo_Q', theo_Q)
# Log iboth model parameters and annealing parameters
dlog.append_all(new_model_params)
dlog.append_all(anneal.as_dict())
return new_model_params
anneal.next()
else:
#======================= Initial model parameters =======================
tracing.tracepoint("Estimate params")
dlog.progress("Estimating initial parameters")
lparams = model.standard_init({'y': my_y})
lparams = model.noisify_params(lparams, anneal=anneal)
if W_init != "estimate": lparams['W'] = W_init
if pi_init != "estimate": lparams['pi'] = pi_init
if sigma_init != "estimate": lparams['sigma'] = sigma_init
#As a control, we also append the initial values
dlog.append_all({
'W': lparams['W'],
'pi': lparams['pi'],
'sigma': lparams['sigma'],
})
comm.Barrier()
#==================== Create and run EM =====================================
dlog.progress("Starting EM")
em = EM(model=model, anneal=anneal)
em.data = {'y': my_y, 'y_rc': my_y_rc}
em.lparams = lparams
em.run(verbose=True)
lparams = em.lparams
#dlog.close()
# Parameters
rf_shape = (26, 26)
H = 16
# Configure Data-Logger
dlog.start_gui(GUI)
dlog.set_handler('W', RFViewer, rf_shape=rf_shape)
dlog.set_handler('S', YTPlotter)
dlog.set_handler('C', YTPlotter)
dlog.set_handler(('T', 'S', 'C'), TextPrinter)
# And GO!
D = rf_shape[0] * rf_shape[1]
Wshape = (H, D)
i = 0
for T in np.linspace(0., 20, 50):
i = i + 1
pprint("%i th iteration..." % i)
W = np.random.normal(size=Wshape)
dlog.append_all({
'T': T,
'W': W,
'S': np.sin(T),
'C': np.cos(T),
})
dlog.close()
anneal.next()
else:
#======================= Initial model parameters =======================
tracing.tracepoint("Estimate params")
dlog.progress("Estimating initial parameters")
lparams = model.standard_init({'y' : my_y})
lparams = model.noisify_params(lparams, anneal=anneal)
if W_init != "estimate": lparams['W'] = W_init
if pi_init != "estimate": lparams['pi'] = pi_init
if sigma_init != "estimate": lparams['sigma'] = sigma_init
#As a control, we also append the initial values
dlog.append_all( {
'W' : lparams['W'],
'pi': lparams['pi'],
'sigma': lparams['sigma'],
} )
comm.Barrier()
#==================== Create and run EM =====================================
dlog.progress("Starting EM")
em = EM(model=model, anneal=anneal)
em.data = {'y': my_y, 'y_rc': my_y_rc}
em.lparams = lparams
em.run(verbose=True)
lparams = em.lparams
#dlog.close()
# Parameters
rf_shape = (26, 26)
H = 16
# Configure Data-Logger
dlog.start_gui(GUI)
dlog.set_handler('W', RFViewer, rf_shape=rf_shape)
dlog.set_handler('S', YTPlotter)
dlog.set_handler('C', YTPlotter)
dlog.set_handler(('T', 'S', 'C'), TextPrinter)
# And GO!
D = rf_shape[0] * rf_shape[1]
Wshape = (H,D)
i = 0
for T in np.linspace(0., 20, 50):
i = i + 1
pprint( "%i th iteration..." % i)
W = np.random.normal(size=Wshape)
dlog.append_all( {
'T': T,
'W': W,
'S': np.sin(T),
'C': np.cos(T),
} )
dlog.close()
