def sample_pos_neg(self, sample_shape=(1, ), eps=None):
if eps is None:
eps = {'W_df': 2 * self.wishart.W.shape[0]}
P_neg, P, P_pos = self.wishart.sample_pos_neg(sample_shape, float(eps))
mus = self.normal.sample(sample_shape)
samples_neg = tl2lt((P_neg, mus))
samples = tl2lt((P, mus))
samples_pos = tl2lt((P_pos, mus))
return samples_neg, samples, samples_pos
def sample_mult_eps(self, sample_shape=(1, ), eps=None):
if eps is None:
eps = {'G_df': 1, 'W_df': 2 * self.wishart.W.shape[0]}
nu_neg, nu, nu_pos = self.gamma.sample_pos_neg(sample_shape,
eps['G_df'])
P_neg, P, P_pos = self.wishart.sample_pos_neg(sample_shape,
eps['W_df'])
samples_neg = {'W_df': tl2lt((nu, P_neg)), 'G_df': tl2lt((nu_neg, P))}
samples_pos = {'W_df': tl2lt((nu, P_pos)), 'G_df': tl2lt((nu_pos, P))}
samples = tl2lt((nu, P))
return samples_neg, samples, samples_pos
def sample_pos_neg(self, sample_shape=(1, ), eps=1):
n = sample_shape[0]
df = self.df.detach().numpy()[0]
mu = self.mu.detach().numpy().reshape(1, self.d)
S = self.S.detach().numpy()
rate = self.rate.detach().numpy()[0]
lam_neg = np.random.gamma(df - eps / 2, 1 / rate, (n, 1))
lam = lam_neg + np.random.gamma(eps / 2, 1 / rate, (n, 1))
lam_pos = lam + np.random.gamma(eps / 2, 1 / rate, (n, 1))
theta = np.random.randn(n, self.d)
scaled_theta = theta.dot(S)
X = mu + scaled_theta
return (tl2lt((cast(X), cast(lam_neg))), tl2lt(
(cast(X), cast(lam))), tl2lt((cast(X), cast(lam_pos))))
def sample(self, sample_shape=(1, )):
n = sample_shape[0]
df = self.df.detach().numpy()[0]
mu = self.mu.detach().numpy().reshape(1, self.d)
S = self.S.detach().numpy()
rate = self.rate.detach().numpy()[0]
lam = np.random.gamma(df, 1 / rate, (n, 1))
theta = np.random.randn(n, self.d)
X = mu + theta.dot(S)
return tl2lt((torch.from_numpy(X.astype(np.float64)),
torch.from_numpy(lam.astype(np.float64))))
def sample(self, sample_shape=(1, )):
P_samples = self.wishart.sample(sample_shape)
mu_samples = self.normal.sample(sample_shape)
return tl2lt((P_samples, mu_samples))
def sample(self, sample_shape=(1, )):
nu_samples = self.gamma.sample(sample_shape)
P_samples = self.wishart.sample(sample_shape)
mu_samples = self.normal.sample(sample_shape)
return tl2lt((nu_samples, P_samples, mu_samples))