calibrar

Noisy sphere test function

Evaluate a stochastic variant of the sphere test function often used in optimisation examples. The deterministic sphere function is the sum of the squared coordinates; here, Gaussian noise is added to each coordinate before the squared distance is computed.

Usage

sphereN(x, sd = 0.1, aggregate = TRUE)

Arguments

• x:

  Numeric vector of coordinates.

• sd:

  Standard deviation of the Gaussian noise added independently to each coordinate.

• aggregate:

  Logical. If TRUE, return the summed squared distance. If FALSE, return the contribution of each coordinate separately.

Value

A numeric scalar if aggregate = TRUE, or a numeric vector of squared contributions if aggregate = FALSE.

Details

This function is mainly useful for examples, testing, and illustrating how calibrar behaves with stochastic objective functions.

See also

calibrate(), optim2(), ahres()

Examples

set.seed(1)
sphereN(rep(0, 5))
#> [1] 0.03777941
sphereN(rep(0, 5), aggregate = FALSE)
#> [1] 0.0067316837 0.0023758708 0.0054512337 0.0033152416 0.0009326207