def test_normalize_multi_factor(self):
values_1 = np.random.rand(5)
values_2 = np.random.rand(4, 3)
values = np.array([values_1, values_2])
d = Dirichlet(values=values)
normed = Categorical(values=d.mean(return_numpy=True))
self.assertTrue(normed.is_normalized())
def test_normalize_two_dim(self):
values = np.array([[1.0, 1.0], [1.0, 1.0]])
d = Dirichlet(values=values)
expected_values = np.array([[0.5, 0.5], [0.5, 0.5]])
self.assertTrue(
np.array_equal(d.mean(return_numpy=True), expected_values))
# Prior We initialise the agent's prior over hidden states at the start to be a flat distribution (i.e. the agent has no strong beliefs about what state it is starting in) # Posterior (recognition density) We initialise the posterior beliefs `qs` about hidden states (namely, beliefs about 'where I am') as a flat distribution over the possible states. This requires the agent to first gather a proprioceptive observation from the environment (e.g. a bodily sensation of where it feels itself to be) before updating its posterior to be centered on the true, evidence-supported location. """ likelihood_matrix = env.get_likelihood_dist() A = Categorical(values=likelihood_matrix) A.remove_zeros() plot_likelihood(A, 'Observation likelihood') b = Dirichlet(values=np.ones((n_states, n_states))) B = b.mean() plot_likelihood(B, 'Initial transition likelihood') D = Categorical(values=np.ones(n_states)) D.normalize() qs = Categorical(dims=[env.n_states]) qs.normalize() """ Run the dynamics of the environment and inference. Start by eliciting one observation from the environment """ # reset environment first_state = env.reset(init_state=s[0]) # get an observation, given the state
