fSolver: User-Specified Loss Function

In isotone: Active Set and Generalized PAVA for Isotone Optimization

User-Specified Loss Function

Description

Specification of a differentiable convex loss function.

Usage

fSolver(z, a, extra)

Arguments

z	Vector containing observed response
a	Matrix with active constraints
extra	List with element fobj containing the target function and gobj with the first derivative

Details

This function is called internally in activeSet by setting mySolver = fSolver. It uses optim() with "BFGS" for optimization.

Value

x	Vector containing the fitted values
lbd	Vector with Lagrange multipliers
f	Value of the target function
gx	Gradient at point x

See Also

activeSet

Examples

##Fitting isotone regression using active set (L2-norm user-specified)
set.seed(12345)
y <- rnorm(9)              ##response values
w <- rep(1,9)              ##unit weights
btota <- cbind(1:8, 2:9)   ##Matrix defining isotonicity (total order)
fit.convex <- activeSet(btota, fSolver, fobj = function(x) sum(w*(x-y)^2), 
gobj = function(x) 2*drop(w*(x-y)), y = y, weights = w)

isotone documentation built on March 7, 2023, 3:17 p.m.