use ndarray::Array1;
use stochastic_rs_core::simd_rng::Deterministic;
use stochastic_rs_core::simd_rng::SeedExt;
use stochastic_rs_core::simd_rng::Unseeded;
use crate::noise::cgns::Cgns;
use crate::traits::FloatExt;
use crate::traits::ProcessExt;
pub struct Cbms<T: FloatExt, S: SeedExt = Unseeded> {
pub rho: T,
pub n: usize,
pub t: Option<T>,
pub seed: S,
cgns: Cgns<T>,
}
impl<T: FloatExt> Cbms<T> {
pub fn new(rho: T, n: usize, t: Option<T>) -> Self {
assert!(
(-T::one()..=T::one()).contains(&rho),
"Correlation coefficient must be in [-1, 1]"
);
Self {
rho,
n,
t,
seed: Unseeded,
cgns: Cgns::new(rho, n - 1, t),
}
}
}
impl<T: FloatExt> Cbms<T, Deterministic> {
pub fn seeded(rho: T, n: usize, t: Option<T>, seed: u64) -> Self {
assert!(
(-T::one()..=T::one()).contains(&rho),
"Correlation coefficient must be in [-1, 1]"
);
Self {
rho,
n,
t,
seed: Deterministic::new(seed),
cgns: Cgns::new(rho, n - 1, t),
}
}
}
impl<T: FloatExt, S: SeedExt> Cbms<T, S> {
#[inline]
fn cumsum_noise(&self, noise: [Array1<T>; 2]) -> [Array1<T>; 2] {
let [cgn1, cgn2] = &noise;
let mut bm1 = Array1::<T>::zeros(self.n);
let mut bm2 = Array1::<T>::zeros(self.n);
for i in 1..self.n {
bm1[i] = bm1[i - 1] + cgn1[i - 1];
bm2[i] = bm2[i - 1] + cgn2[i - 1];
}
[bm1, bm2]
}
}
impl<T: FloatExt, S: SeedExt> ProcessExt<T> for Cbms<T, S> {
type Output = [Array1<T>; 2];
fn sample(&self) -> Self::Output {
let noise = self.cgns.sample_impl(&self.seed.derive());
self.cumsum_noise(noise)
}
}
py_process_2x1d!(PyCbms, Cbms,
sig: (rho, n, t=None, seed=None, dtype=None),
params: (rho: f64, n: usize, t: Option<f64>)
);