hessApprox: Numerical approximation of Hessian

In thiyangt/fformpp: Feature-based FORecast Model Performance Prediction

Numerical approximation of Hessian

Description

Approximation of hessian matrix by using different methods. The gradient should be always a *matrix*. Current method is "outer" for outer product. If the gradient is not an one-column matrix, it will convert it to an one-column matrix.

Usage

hessApprox(gradient, method)

Arguments

gradient	"matrix". p-by-q, the gradient vector/matrix.
method	"character". Method to be used in the approximation. Possible value may be "outer" for outer product of gradient.
matrix	"Logical" Should the Hessian matrix be returned or just the diagonal vector.

Note

Created: Tue Mar 30 16:33:23 CEST 2010. Current: Mon Feb 09 18:34:40 CST 2015.

Author(s)

Feng Li, Dept. of Statistics, Stockholm University, Sweden.

References

Villani et al (2009); Li et al(2010a, b); Train, K. 2003, Discrete choice methods with simulation(2ed), Cambridge Univ. Press. p. 193.

thiyangt/fformpp documentation built on Jan. 5, 2024, 5:44 a.m.