I am reading few papers on Sparse Bayesian Learning topic. Many papers use posterior calculation or approximation using likelihood function and priors. Usually priors are defined like this P(w|Alpha) along with some hyper parameter. Furthermore, some techniques use flat Gamma Hyperpriors together with these priors.
I want to know, what do we initialize at the start of algorithm. We have unknown parameters to be initialized. What are these parameters which are to be initialized keeping in view the Priors particularly.