For neural networks, Pyro can set random weights for a single layer. My question is that is there an approach to set random weights for only individual weight or single weight other than all of the weights in the same layer?
You can set prior for one specific weight vector, but it seems to be tricky to set prior for only one element of that weight vector. Probably you need to split your weight vector into two parts: one to be set prior, and one to be an nn parameter.