`MLPKriging::simulate`

Description

Simulate paths from an MLPKriging model.

Usage

Python

# mk = MLPKriging(...)
mk.simulate(nsim = 1, seed = 123, x = x, will_update = False)

R

# mk <- MLPKriging(...)
mk$simulate(nsim = 1, seed = 123, x = x, will_update = FALSE)

Matlab/Octave

% mk = MLPKriging(...)
mk.simulate(nsim = 1, seed = 123, x = x, will_update = false)

Julia

# mk = MLPKriging(...)
s = simulate(mk, nsim=1, seed=123, x=x, will_update=false)

Arguments

Argument	Description
`nsim`	Number of simulation paths. Default `1`.
`seed`	Random seed. Default `123`.
`x`	Numeric matrix of simulation points (\(m \times d\)).
`will_update`	Logical. Set to `TRUE` if `update_simulate` will be called afterwards. Default `FALSE`.

Details

The MLP feature extractor is applied first, then the GP conditional simulation is performed in the learned latent feature space.

Value

A matrix with nrow(x) rows and nsim columns containing the simulated paths.

Examples

f <- function(x) 1 - 1 / 2 * (sin(12 * x) / (1 + x) + 2 * cos(7 * x) * x^5 + 0.7)
X <- as.matrix(seq(0.05, 0.95, length.out = 10))
y <- f(X)

mk <- MLPKriging(
  y, X,
  hidden_dims = c(4L),
  d_out = 1L,
  activation = "tanh",
  kernel = "gauss",
  parameters = list(max_iter_adam = "20", max_iter_bfgs = "10")
)
x <- as.matrix(seq(0, 1, length.out = 101))
s <- mk$simulate(nsim = 10, seed = 123, x = x)

plot(f)
points(X, y, col = "blue")
matlines(x, s, col = rgb(0, 0, 1, 0.2), type = "l", lty = 1)

MLPKriging::simulate