Missing labels in Bayesian neural network

[Prasad Deshpande]

Hi,

I am working on a binary classification of a time-series data.

Some of the labels (targets) are missing.

I am planning for probabilistic Bayesian neural network similar to below link https://keras.io/examples/keras_recipes/bayesian_neural_networks/

Please let me know how to incorporate the missing values in the same. I believe the loss function needs to be modified. But I am confused how to modify the negative log likelihood loss function.

Also let me know if there can be any other way also.

Thanks in advance,

Prasad