I am using audio functions having the form (1368, 64) to learn Bayesian network structure with pgmpy. The actual labels of the functions are also added in the data frame. But after learning, the BN does not include the output node, which prevents it from making inferences. What I can do? Can I bind the output label node to the feature nodes randomly or try other methods?

I tried changing the structure learning algorithm. I want to know if I can link the output node to the nodes of the Bayesian network randomly.