1. In the paper, the author mentioned maximizing the sum of Q(θ|θold) and the log-prior probability by deriving the Laplace scale parameter ri. I was wondering if someone could provide some insights into the methodology or approach that used to derive this parameter. I am particularly interested in the mathematical details behind its determination.

2. I am also curious about the potential applicability of the learning algorithm described in your paper to the nonlinear case. If it is possible to generalize this algorithm to nonlinear systems, could someone please share any insights on how the parameters C and r should be estimated in such cases?