def get_sample(self, number_samples, input_values={}): #TODO: code duplication here
reformatted_input_values = reformat_sampler_input(pandas_frame2dict(input_values),
number_samples=number_samples)
raw_sample = self._get_sample(number_samples, observed=False, input_values=reformatted_input_values,
differentiable=False)
sample = reformat_sample_to_pandas(raw_sample)
return sample
def get_posterior_sample(self, number_samples, input_values={}):
reformatted_input_values = reformat_sampler_input(
pandas_frame2dict(input_values), number_samples=number_samples)
raw_sample = self._get_posterior_sample(
number_samples, input_values=reformatted_input_values)
sample = reformat_sample_to_pandas(raw_sample)
return sample
def get_sample(self, number_samples, input_values={}):
reformatted_input_values = reformat_sampler_input(pandas_frame2dict(input_values),
number_samples=number_samples)
raw_sample = {self: self._get_sample(number_samples, resample=False,
observed=self.is_observed,
differentiable=False,
input_values=reformatted_input_values)[self]}
sample = reformat_sample_to_pandas(raw_sample)
self.reset()
return sample
def get_entropy(self, input_values={}):
reformatted_input_values = reformat_sampler_input(
pandas_frame2dict(input_values), number_samples=1)
raw_ent = self._get_entropy(reformatted_input_values)
ent = reformat_sample_to_pandas(raw_ent, number_samples=1)
return ent
def get_variance(self, input_values={}):
reformatted_input_values = reformat_sampler_input(
pandas_frame2dict(input_values), number_samples=1)
raw_variance = self._get_variance(reformatted_input_values)
variance = reformat_sample_to_pandas(raw_variance, number_samples=1)
return variance
def get_mean(self, input_values={}):
reformatted_input_values = reformat_sampler_input(
pandas_frame2dict(input_values), number_samples=1)
raw_mean = self._get_mean(reformatted_input_values)
mean = reformat_sample_to_pandas(raw_mean, number_samples=1)
return mean
biased=False,
number_post_samples=8000000)
inference.perform_inference(model,
inference_method=inference_method,
number_iterations=1000,
number_samples=50,
optimizer="SGD",
lr=0.0001,
posterior_model=particles,
pretraining_iterations=0)
loss_list = model.diagnostics["loss curve"]
# Local variational models
plt.plot(loss_list)
plt.show()
# Samples
print(inference_method.weights)
M = 8000
samples = [reformat_sample_to_pandas(sampler._get_sample(M, max_itr=np.inf)) for sampler in inference_method.sampler_model]
ensemble_histogram(samples,
variable="theta",
weights=inference_method.weights,
bins=50)
plt.show()
#print([p.get_sample(1) for p in particles])
#print(initial_locations)
#samples = Qtheta.get_sample(50)
#print(samples)
