EAM.converged: Check convergence of the EAM algorithm.

In depCensoring: Statistical Methods for Survival Data with Dependent Censoring

Check convergence of the EAM algorithm.

Description

This function checks the convergence of the EAM algorithm. ToDo: Get rid of hard coding stop of algorithm when no more improvement of theta values (maybe related to parameter space contraction, since the problem is that the given points to check in the E-step of the following iteration can always be the same and always be rejected (except of course for the randomly chosen one), while the most promising theta value continues to be the same, infeasible value. In this way, it is possible that theta.hash - mp.theta.next at some point will never decrease).

Usage

EAM.converged(
  opt.val.prev,
  evaluations,
  mp.theta.next,
  iter.nbr,
  dir,
  hyperparams,
  verbose
)

Arguments

opt.val.prev	Previous optimal theta value.
evaluations	Matrix of violation curve evaluations.
mp.theta.next	Most promising value of theta for which to run the E-step in the following iteration
iter.nbr	Number of iterations of the EAM algorithm run so far.
dir	Search direction.
hyperparams	List of hyperparameters used in the EAM algorithm.
verbose	Verbosity parameter.

Value

Boolean value whether or not algorithm has converged.

depCensoring documentation built on April 4, 2025, 1:52 a.m.