From: Multi-dimensional model order selection
Approach
Penalty function p(P, N, α(G))
R-D AIC
P · (2 · α(G) - P)
R-D MDL