I built a Bayesian neural network and the loss value is smaller than 0, but the prediction result is reasonable. Is this normal?
Thank you, the results are indeed reasonable.
