Is there a possibility to calculate the variance of target values for each batch using keras custom callback?
Maybe in a similar manner to the Create keras callback to save model predictions and targets for each batch during training ? Im using keras from tensorflow backend and python3.
Cheers,
Maks