numderiv: Numerical (approximate) Differentiation.

In madness: Automatic Differentiation of Multivariate Operations

Numerical (approximate) Differentiation.

Description

Approximates the derivative of a function at the input by numerical methods.

Usage

numderiv(f, x, eps=1e-8, type=c('forward','central','backward'),...)

## S4 method for signature 'ANY,array'
numderiv(f, x, eps = 1e-08, type = c("forward", "central", "backward"), ...)

## S4 method for signature 'ANY,madness'
numderiv(f, x, eps = 1e-08, type = c("forward", "central", "backward"), ...)

Arguments

f	a function, to be evaluated at and near x.
x	array, matrix, or madness object.
eps	the 'epsilon', a small value added or subtracted from x to compute the first differences.
type	the type of first difference, case-insensitive, substrings ok.
...	arguments passed on to f.

Details

For a multivariate-valued function of multivariate data, approximates the derivative at a point via the forward, central, or backward first differences, returning a madness object.

Value

A matrix if x is numeric; a madness object if x is a madness object.

Author(s)

Steven E. Pav [e-mail]

Examples

f <- function(x,h) {
  cos(x + h)
}
x <- array(rnorm(100),dim=c(10,10))
madx <- numderiv(f,x,1e-8,h=pi)

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