onload: Default Options for lavaPenalty
In bozenne/lava.penalty: Penalized Latent Variable Models

Default options for lavaPenalty.

1
2
3

.onLoad(lib, pkg = "lavaPenalty")

.onAttach(lib, pkg = "lavaPenalty")

Add the following elements to lava.options:

item iter.max maximum number of iterations item trace should the convergence diagnostics be displayed at each step item abs.tol convergence is the difference in likelihood between two consecutive steps is below this threshold item rel.tol convergence is the relative difference in likelihood between two consecutive steps is below this threshold item step maximun step for the proximal gradient algorithm. If NULL the step is estimated using the inverse of the maximal eigenvalue of the hessian (in absolute value) and re-estimated at each step Otherwise backtracking is used. item BT.n number of backtracking steps item BT.eta multiplicative factor for the step item force.descent item export.iter should all iterations be exported (details.cv) item method type of iteration ISTA correspond to the ISTA step as described in Bech 2009 FISTA_Beck correspond to the FISTA step as described in Bech 2009 FISTA_Vand correspond to ?? mFISTA_Vand correspond to ?? item resolution_lambda1 the first value is the maximum relative difference in parameter between two steps. If this lead to a too small step, the second value is used as the minimum change in penalization parameter between two steps. item increasing direction of the path item stopLambda if not null, stop the path when the penalty parameter reach this value item stopParam if not null, stop the path when the number of 0 (if increasing = TRUE) or non 0 (if increasing = FALSE) parameters has reached this value. item nstep_max the maximum number of iterations item ode.method the type of method to use to solve the ode (see the documentation of deSolve:::ode) item constrain the constrain on the variance parameters item reversible should the algorithm allow a 0 parameter to become non-0 (when increasing = TRUE) or non-0 parameter to become 0 (when increasing = FALSE) item tol.0 tolerance for classify a parameter from beta_lambdaMax in the set of 0 parameters item exportAllPath export all the regularization path (and not only the breakpoints) item trace should the function be traced

bozenne/lava.penalty documentation built on May 13, 2019, 1:41 a.m.