egobox_doe/
traits.rs

1use linfa::Float;
2use ndarray::Array2;
3
4/// Sampling method allowing to generate a DoE in a given sample space
5///
6/// A sampling method is able to generate a set of `ns` samples in a given sample space.
7/// where the sample space is defined by `[lower_bound_xi, upper_bound_xi]^nx`
8/// within `R^nx` where `nx` is the dimension of the sample space: x = (x_i) with i in [1, nx].
9pub trait SamplingMethod<F: Float> {
10    /// Returns the bounds of the sample space
11    ///
12    /// # Returns
13    ///
14    /// * A (nx, 2) matrix where the ith row is the interval of the ith components of a sample.
15    fn sampling_space(&self) -> &Array2<F>;
16
17    /// Generates a (ns, nx)-shaped array of samples belonging to `[0., 1.]^nx`
18    ///
19    /// # Parameters
20    ///
21    /// * `ns`: number of samples
22    ///
23    /// # Returns
24    ///
25    /// * A (ns, nx) matrix of samples where nx is the dimension of the sample space
26    ///   each sample belongs to `[0., 1.]^nx` hypercube
27    fn normalized_sample(&self, ns: usize) -> Array2<F>;
28
29    /// Generates a (ns, nx)-shaped array of samples belonging to `[lower_bound_xi, upper_bound_xi]^nx`
30    ///
31    /// # Parameters
32    ///
33    /// * `ns`: number of samples
34    ///
35    /// # Returns
36    ///
37    /// * A (ns, nx) matrix where nx is the dimension of the sample space.
38    ///   each sample belongs to `[lower_bound_xi, upper_bound_xi]^nx` where bounds
39    ///   are defined as returned values of `sampling_space` function.
40    fn sample(&self, ns: usize) -> Array2<F> {
41        let xlimits = self.sampling_space();
42        let lower = xlimits.column(0);
43        let scaler = &xlimits.column(1) - &lower;
44        self.normalized_sample(ns) * scaler + lower
45    }
46}
egobox_doe/traits.rs

egobox_doe/
traits.rs
