def sample_pp_post_pred(self, nsamp, input_list, full_cov=False, nloop=None):
""" Sample from posterior predictive of PP.
Inputs:
input_list - list of np arrays size=(-1,)
Returns:
list (len input_list) of np arrays (size=(nsamp,1))."""
if self.modelp.model_str=='optfixedsig' or self.modelp.model_str=='opt' or \
self.modelp.model_str=='optmatern32':
nloop = 1
sampids = [0]
elif self.modelp.model_str=='samp' or \
self.modelp.model_str=='sampmatern32':
if nloop is None: nloop=nsamp
nsamp = int(nsamp/nloop)
sampids = np.random.randint(len(self.sample_list), size=(nloop,))
ppred_list = []
for i in range(nloop):
samp = self.sample_list[sampids[i]]
postmu, postcov = self.gp_post(self.data.X, self.data.y,
np.stack(input_list), samp.ls, samp.alpha, samp.sigma, full_cov)
if full_cov:
ppred_list.extend(list(sample_mvn(postmu, postcov, nsamp)))
else:
ppred_list.extend(list(np.random.normal(postmu.reshape(-1,),
postcov.reshape(-1,), size=(nsamp, len(input_list)))))
return self.get_reverse_transform(list(np.stack(ppred_list).T), ppred_list,
input_list)
def sample_pp_pred(self, nsamp, input_list, lv=None):
""" Sample from predictive of PP for parameter lv.
Returns: list (len input_list) of np arrays (size (nsamp,1))."""
if lv is None:
lv = self.sample_list[0]
postmu, postcov = self.gp_post(self.data.X, self.data.y, input_list,
lv.ls, lv.alpha, lv.sigma)
pred_list = list(sample_mvn(
postmu, postcov, 1)) ###TODO: sample from this mean nsamp times
return list(np.stack(pred_list).T), pred_list
def sample_pp_pred(self, nsamp, input_list, lv=None):
""" Sample from predictive of PP for parameter lv.
Returns: list (len input_list) of np arrays (size (nsamp,1))."""
x_pred = np.stack(input_list)
if lv is None:
if self.modelp.model_str=='optfixedsig' or self.modelp.model_str=='opt' \
or self.modelp.model_str=='fixedparam':
lv = self.sample_list[0]
elif self.modelp.model_str=='samp' or \
self.modelp.model_str=='sampfixedsig':
lv = self.sample_list[np.random.randint(len(self.sample_list))]
postmu, postcov = self.gp_post(self.data.X, self.data.y, x_pred, lv.ls,
lv.alpha, lv.sigma)
pred_list = list(sample_mvn(postmu, postcov, 1)) ###TODO: sample from this mean nsamp times
return list(np.stack(pred_list).T), pred_list
def sample_pp_post_pred(self, nsamp, input_list, full_cov=False):
""" Sample from posterior predictive of PP.
Inputs:
input_list - list of np arrays size=(-1,)
Returns:
list (len input_list) of np arrays (size=(nsamp,1))."""
samp = self.sample_list[0]
postmu, postcov = self.gp_post(self.data.X, self.data.y, input_list,
samp.ls, samp.alpha, samp.sigma,
full_cov)
if full_cov:
ppred_list = list(sample_mvn(postmu, postcov, nsamp))
else:
ppred_list = list(
np.random.normal(postmu.reshape(-1, ),
postcov.reshape(-1, ),
size=(nsamp, len(input_list))))
return list(np.stack(ppred_list).T), ppred_list