fdars-core 0.17.0

Functional Data Analysis algorithms in Rust
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
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148

//! Benchmarks for regression and FPCA methods
//!
//! Compares performance of:
//! - FPCA (fdata_to_pc_1d) with varying n and m
//! - Functional linear regression (fregre_lm) with varying ncomp
//! - Functional logistic regression with varying ncomp

use criterion::{black_box, criterion_group, criterion_main, BenchmarkId, Criterion};
use fdars_core::matrix::FdMatrix;
use fdars_core::regression::fdata_to_pc_1d;
use fdars_core::scalar_on_function::{fregre_lm, functional_logistic};
use std::f64::consts::PI;

/// Generate synthetic functional data (n curves, m time points) and a scalar response.
///
/// The response is a linear combination of the first few FPC scores + noise,
/// so that fregre_lm has something meaningful to fit.
fn generate_regression_data(n: usize, m: usize) -> (FdMatrix, Vec<f64>, Vec<f64>) {
    let argvals: Vec<f64> = (0..m).map(|j| j as f64 / (m - 1) as f64).collect();
    let mut data = vec![0.0; n * m];

    for i in 0..n {
        let c1 = ((i as f64 * 3.7).sin()) * 2.0;
        let c2 = ((i as f64 * 5.1 + 0.3).sin()) * 1.5;
        let c3 = ((i as f64 * 7.9 + 0.7).sin()) * 1.0;
        for j in 0..m {
            let t = argvals[j];
            data[i + j * n] =
                c1 * (2.0 * PI * t).sin() + c2 * (4.0 * PI * t).sin() + c3 * (6.0 * PI * t).sin();
        }
    }

    // Response is a function of the underlying coefficients
    let y: Vec<f64> = (0..n)
        .map(|i| {
            let c1 = ((i as f64 * 3.7).sin()) * 2.0;
            let c2 = ((i as f64 * 5.1 + 0.3).sin()) * 1.5;
            let noise = ((i as f64 * 13.7).sin()) * 0.1;
            2.0 * c1 + 1.0 * c2 + noise
        })
        .collect();

    let mat = FdMatrix::from_column_major(data, n, m).unwrap();
    (mat, y, argvals)
}

/// Generate binary response data for logistic regression.
fn generate_logistic_data(n: usize, m: usize) -> (FdMatrix, Vec<f64>, Vec<f64>) {
    let (mat, y_continuous, argvals) = generate_regression_data(n, m);
    // Convert to binary based on sign of y
    let y_binary: Vec<f64> = y_continuous
        .iter()
        .map(|&yi| if yi >= 0.0 { 1.0 } else { 0.0 })
        .collect();
    (mat, y_binary, argvals)
}

fn bench_fpca(c: &mut Criterion) {
    let mut group = c.benchmark_group("fdata_to_pc_1d");

    for &n in &[50, 200] {
        for &m in &[50, 100] {
            let (data, _y, argvals) = generate_regression_data(n, m);
            let ncomp = 5;
            let label = format!("n{}_m{}", n, m);

            group.bench_with_input(BenchmarkId::new("params", &label), &label, |b, _| {
                b.iter(|| fdata_to_pc_1d(black_box(&data), black_box(ncomp), black_box(&argvals)));
            });
        }
    }

    group.finish();
}

fn bench_fregre_lm(c: &mut Criterion) {
    let mut group = c.benchmark_group("fregre_lm");
    let n = 100;
    let m = 50;
    let (data, y, _argvals) = generate_regression_data(n, m);

    for &ncomp in &[3, 5, 10] {
        group.bench_with_input(BenchmarkId::new("ncomp", ncomp), &ncomp, |b, &ncomp| {
            b.iter(|| fregre_lm(black_box(&data), black_box(&y), None, black_box(ncomp)));
        });
    }

    // Also vary n with fixed ncomp
    for &n in &[50, 200] {
        let (data_n, y_n, _) = generate_regression_data(n, m);
        let label = format!("n{}_nc5", n);
        group.bench_with_input(BenchmarkId::new("scale", &label), &label, |b, _| {
            b.iter(|| fregre_lm(black_box(&data_n), black_box(&y_n), None, black_box(5)));
        });
    }

    group.finish();
}

fn bench_functional_logistic(c: &mut Criterion) {
    let mut group = c.benchmark_group("functional_logistic");
    let n = 100;
    let m = 50;
    let (data, y, _argvals) = generate_logistic_data(n, m);

    for &ncomp in &[3, 5, 10] {
        group.bench_with_input(BenchmarkId::new("ncomp", ncomp), &ncomp, |b, &ncomp| {
            b.iter(|| {
                functional_logistic(
                    black_box(&data),
                    black_box(&y),
                    None,
                    black_box(ncomp),
                    black_box(100),
                    black_box(1e-6),
                )
            });
        });
    }

    // Vary n with fixed ncomp
    for &n in &[50, 200] {
        let (data_n, y_n, _) = generate_logistic_data(n, m);
        let label = format!("n{}_nc5", n);
        group.bench_with_input(BenchmarkId::new("scale", &label), &label, |b, _| {
            b.iter(|| {
                functional_logistic(
                    black_box(&data_n),
                    black_box(&y_n),
                    None,
                    black_box(5),
                    black_box(100),
                    black_box(1e-6),
                )
            });
        });
    }

    group.finish();
}

criterion_group!(
    benches,
    bench_fpca,
    bench_fregre_lm,
    bench_functional_logistic
);
criterion_main!(benches);