return s_1_t, s_2_t, s_3_t, theta_b_t, theta_c_t, theta_sig_t, v_t
((s_1_t, s_2_t, s_3_t, theta_b_t, theta_c_t, theta_sig_t, v_t), updates) = theano.scan(fn=inner_fn,
sequences=[x],
outputs_info=[s_1_0, s_2_0, s_3_0, None, None, None, None])
for k, v in updates.iteritems():
k.default_update = v
lstm_1_tm1 = s_1_t[-1]
lstm_2_tm1 = s_2_t[-1]
lstm_3_tm1 = s_3_t[-1]
theta_b_in = theta_b_t.reshape((x_shape[0]*x_shape[1], -1))
theta_c_in = theta_c_t.reshape((x_shape[0]*x_shape[1], -1))
theta_sig_in = theta_sig.fprop()
v_in = v_t.reshape((x_shape[0]*x_shape[1], -1))
W = grbm.params['W_x__grbm']
free_energy = grbm_free_energy(x_in, W, [theta_b_in, theta_c_in, theta_sig_in]) -\
grbm_free_energy(v_in, W, [theta_b_in, theta_c_in, theta_sig_in])
free_energy = free_energy.mean()
free_energy.name = 'grbm_free_energy'
recon_err = T.sqr(x_in - v_in).mean()
recon_err.name = 'reconstruction_error'
max_x = x.max()
mean_x = x.mean()
min_x = x.min()
max_x.name = 'max_x'
mean_x.name = 'mean_x'
marginal_ll = []
for i in xrange(num_sample):
z_is = prior.fprop([phi_mu_t, phi_sig_t])
theta_emb_t = theta_emb.fprop([dec_t])
theta_mu_t = theta_mu.fprop([theta_emb_t])
theta_sig_t = theta_sig.fprop([theta_emb_t])
coeff_t = coeff.fprop([theta_emb_t])
w = Gaussian(z_is, prior_mu_t, prior_sig_t) -\
Gaussian(z_is, phi_mu_t, phi_sig_t)
marginal_ll.append(GMM(x_t, theta_mu_t, theta_sig_t, coeff_t) + w)
marginal_ll = T.concatenate(marginal_ll, axis=0).mean()
return enc_t, dec_t, pec_t, kl_t, theta_mu_t, coeff_t, marginal_ll
prior_sig_t = prior_sig.fprop()
phi_sig_t = phi_sig.fprop()
theta_sig_t = theta_sig.fprop()
((enc_t, dec_t, pec_t, kl_t, theta_mu_t, coeff_t, marginal_ll), updates) =\
theano.scan(fn=inner_fn,
sequences=[x, x2],
outputs_info=[encoder.get_init_state(),
decoder.get_init_state(),
pecoder.get_init_state(),
None, None, None, None],
non_sequences=[phi_sig_t, prior_sig_t, theta_sig_t])
for k, v in updates.iteritems():
k.default_update = v
reshaped_x = x.reshape((x.shape[0]*x.shape[1], -1))
reshaped_theta_mu = theta_mu_t.reshape((theta_mu_t.shape[0]*theta_mu_t.shape[1], -1))