helena 0.1.0

Core types and component interfaces for helena, a latent data-to-waveform generation platform.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40

use super::*;

#[test]
fn fit_recovers_known_statistics() {
    let latent = seq(3, 2, vec![0.0, 1.0, 2.0, 1.0, 4.0, 1.0]);
    let norm = LatentNorm::fit([&latent], LatentNorm::DEFAULT_EPS).unwrap();
    assert_eq!(norm.dim(), 2);
    assert!((norm.mean()[0] - 2.0).abs() < 1e-6);
    assert!((norm.mean()[1] - 1.0).abs() < 1e-6);
    assert!((norm.std()[0] as f64 - (8.0f64 / 3.0).sqrt()).abs() < 1e-5);
    // A constant dimension floors at eps.
    assert_eq!(norm.std()[1], LatentNorm::DEFAULT_EPS);
}

#[test]
fn fit_pools_across_latents() {
    let a = seq(2, 1, vec![0.0, 2.0]);
    let b = seq(2, 1, vec![4.0, 6.0]);
    let norm = LatentNorm::fit([&a, &b], LatentNorm::DEFAULT_EPS).unwrap();
    assert_eq!(norm.dim(), 1);
    assert!((norm.mean()[0] - 3.0).abs() < 1e-6);
}

#[test]
fn standardize_centers_and_scales_the_fit_data() {
    let latent = seq(3, 2, vec![0.0, 10.0, 2.0, 20.0, 4.0, 30.0]);
    let norm = LatentNorm::fit([&latent], LatentNorm::DEFAULT_EPS).unwrap();
    let std = norm.standardize(&latent).unwrap();
    for d in 0..2 {
        let col: Vec<f64> = (0..3).map(|f| std.values()[f * 2 + d] as f64).collect();
        let mean = col.iter().sum::<f64>() / 3.0;
        let var = col.iter().map(|x| (x - mean).powi(2)).sum::<f64>() / 3.0;
        assert!(mean.abs() < 1e-5, "dim {d} mean {mean}");
        assert!(
            (var.sqrt() - 1.0).abs() < 1e-4,
            "dim {d} std {}",
            var.sqrt()
        );
    }
}