Derivatives of the Log of a Determinant
摘要
We present an efficient way to calculate effective model dimensions, using automated differentiation of the Cholesky algorithm. The method is illustrated with two examples using P-splines: adaptive smoothing and smoothing of over-dispersed counts.