| Bisect {Icens} | R Documentation |
Most of the optimizations in Icens have a one dimensional root-finding component. Since the quantities involved are generally restricted to a subset of [0,1] we use bisection to find the roots.
Bisect(tA, pvec, ndir, Meps, tolbis=1e-07)
tA |
The transpose of the clique matrix. |
pvec |
The current estimate of the probability vector. |
ndir |
The direction to explore. |
Meps |
Machine epsilon, elements of pvec that are less
than this are assumed to be zero. |
tolbis |
The tolerance used to determine if the algorithm has converged. |
We search from pvec in the direction ndir to obtain the
new value of pvec that maximizes the likelihood.
The new estimate of pvec.
Alain Vandal and Robert Gentleman.
Any book on optimization.