Create a Random State

Creates a random state of its own, which the samplers draw from in place of the global random state within with_nv_rng_state() or after local_nv_rng_state(). Draws advance it in place. A function compiled with jit() can draw from a random state it takes as an argument or closes over, but cannot create or return one.

See the Random Number Generation article for more details.

Usage

nv_rng_state(seed)

Arguments

seed

(integer(1))
The seed, a 32-bit integer, which gives the same random state as nv_set_seed() gives the global random state.

Value

(AnvlRngState)
A random state of the key that seed gives and the counter 0.

See also

Other rng: local_nv_rng_state(), nv_normal, nv_rbinom(), nv_sample(), nv_sample_int(), nv_set_seed(), nv_uniform

Examples

rng <- nv_rng_state(42L)
with_nv_rng_state(rng, nv_runif(3L))
#> AnvlArray
#>  0.5303
#>  0.3134
#>  0.9015
#> [ CPUf32{3} ] 
# the random state was advanced, so the next draw differs
with_nv_rng_state(rng, nv_runif(3L))
#> AnvlArray
#>  0.7277
#>  0.7879
#>  0.1817
#> [ CPUf32{3} ] 
rng
#> <AnvlRngState> key 0x000000000000002a, counter 0x0000000000000002

# a jitted function takes a random state as an argument
f <- jit(function(x, rng) with_nv_rng_state(rng, x + nv_rnorm(2L)))
f(nv_array(c(1, 2)), nv_rng_state(1L))
#> AnvlArray
#>  1.8863
#>  0.7194
#> [ CPUf32{2} ]