@yyaodong
Hi Dustin:
Let’s say you want to build a predictive model in a Bayesian way, to predict
y_i
based onx_i
, and in the graph definitionP(y_i, x_i, z_i, beta),
there is a local latent variablez_i
that is dependent onx_i
and global parameterbeta
, andy_i
is dependent on all the rest.
I want to know, in this case, after making inference on
z_i
andbeta
based on the training data, how to make predictions when the test datax_test
comes in. Sincez_i
is dependent onx_i
, how can I updatez_i
to thez_test
? As in my understanding, the approximatorq_z_i
is only valid on the training data set, therefore, you cannot simplyq_z_i.copy()
, and in GMM, thez_i
can just be easily reset byqz_init.run()
, becausez_i
is not dependent onx_i
.
Many thanks,