kyle-dorman/bayesian-neural-network-blogpost

Evaluating variance in output

pengnam opened this issue · 1 comments

image

The model trains with both logits_variance and softmax as the output.

softmax is trained with the classification value

What is logits variance trained with?

@pengnam Its kind of a unsupervised learning. We use Monte carlo simulation to produce estimated true values for the variance of any given input and the prediction of NN is trained with those values.