to_objective: Convert a madness object into an objective value with...

In madness: Automatic Differentiation of Multivariate Operations

Convert a madness object into an objective value with gradient

Description

Given a madness object representing a scalar value, strip out that value and attach an attribute of its derivative as a gradient. This is a convenience method that simplifies construction of objective functions for optimization routines.

Usage

to_objective(X)

Arguments

X	a madness object representing a scalar.

Value

A scalar numeric with a gradient attribute of the derivative.

Note

An error will be thrown if the value is not a scalar.

Author(s)

Steven E. Pav [e-mail]

Examples

# an objective function for matrix factorization with penalty:
fitfun <- function(R,L,Y,nu=-0.1) {
 dim(R) <- c(length(R),1)
Rmad <- madness(R)
dim(Rmad) <- c(ncol(L),ncol(Y))
Err <- Y - L %*% Rmad
penalty <- sum(exp(nu * Rmad))
fit <- norm(Err,'f') + penalty
to_objective(fit)
}
set.seed(1234)
L <- array(runif(30*5),dim=c(30,5)) 
Y <- array(runif(nrow(L)*20),dim=c(nrow(L),20))
R0 <- array(runif(ncol(L)*ncol(Y)),dim=c(ncol(L),ncol(Y)))
obj0 <- fitfun(R0,L,Y)
fooz <- nlm(fitfun, R0, L, Y, iterlim=3)

madness documentation built on Aug. 21, 2023, 9:07 a.m.