nr_iterate {lgspline} | R Documentation |
Compute Newton-Raphson Parameter Update with Numerical Stabilization
Description
Performs parameter update in iterative optimization.
Called by damped_newton_r
in the update step
Usage
nr_iterate(gradient_val, neghessian_val)
Arguments
gradient_val |
Numeric vector of gradient values ( |
neghessian_val |
Negative Hessian matrix ( |
Details
This helper function is a core component of Newton-Raphson optimization.
It provides a computationally-stable approach to computing \textbf{G}\textbf{u}
, for
information matrix \textbf{G}
and score vector \textbf{u}
, where the Newton-Raphson update
can be expressed as \boldsymbol{\beta}^{(m+1)} = \boldsymbol{\beta}^{(m)} + \textbf{G}\textbf{u}
.
Value
Numeric vector of parameter updates (\textbf{G}\textbf{u}
)
See Also
damped_newton_r
for the full optimization routine
[Package lgspline version 0.2.0 Index]