Question about AutoDelta guide

I have a quick question about the AutoDelta guide.
If in Bayesian neural network, if I specify my guide for the weights as the AutoDelta guide, would it have the same effect as assigning one identical value to all weights of the Bayesian neural network, where that identical value is drawn from the trained delta posterior (which has probability of 1 for a single point and the probability of 0 for all other values)?

Or would the value for each weight that are drawn from the trained delta posterior be different, since the parameter of the delta posteriors for each weight are different?

Thank you,

Hi @h56cho, I believe it is the second behavior.