fdars_core/depth/
mod.rs

1//! Depth measures for functional data.
2//!
3//! This module provides various depth measures for assessing the centrality
4//! of functional observations within a reference sample.
5
6use crate::matrix::FdMatrix;
7use crate::maybe_par_chunks_mut_enumerate;
8use rand::prelude::*;
9use rand_distr::StandardNormal;
10
11pub mod band;
12pub mod dispatch;
13pub mod erl;
14pub mod extremal;
15pub mod fraiman_muniz;
16pub mod half_region;
17pub mod hypo_epi;
18pub mod linf;
19pub mod modal;
20pub mod random_projection;
21pub mod random_tukey;
22pub mod rpd;
23pub mod spatial;
24pub mod tvd;
25
26#[cfg(test)]
27mod tests;
28
29// Re-export all public functions
30pub use band::{band_1d, modified_band_1d, modified_epigraph_index_1d};
31pub use dispatch::{functional_boxplot, functional_depth, DepthMethod, FunctionalBoxplotResult};
32pub use erl::extreme_rank_length_depth_1d;
33pub use extremal::extremal_depth_1d;
34pub use fraiman_muniz::{fraiman_muniz_1d, fraiman_muniz_2d};
35pub use half_region::{half_region_depth_1d, modified_half_region_depth_1d};
36pub use hypo_epi::{epigraph_index_1d, hypograph_index_1d, modified_hypograph_index_1d};
37pub use linf::linfinity_depth_1d;
38pub use modal::{modal_1d, modal_2d};
39pub use random_projection::{
40    random_projection_1d, random_projection_1d_seeded, random_projection_2d,
41};
42pub use random_tukey::{random_tukey_1d, random_tukey_1d_seeded, random_tukey_2d};
43pub use rpd::{rpd_depth_1d, rpd_depth_1d_seeded};
44pub use spatial::{
45    functional_spatial_1d, functional_spatial_2d, kernel_functional_spatial_1d,
46    kernel_functional_spatial_2d,
47};
48pub use tvd::{total_variation_depth_1d, TvdMssResult};
49
50// ---------------------------------------------------------------------------
51// Shared helpers
52// ---------------------------------------------------------------------------
53
54/// Generate `nproj` unit-norm random projection vectors of dimension `m`.
55///
56/// Returns a flat buffer of length `nproj * m` where projection `p` occupies
57/// `[p*m .. (p+1)*m]`.
58///
59/// If `seed` is Some, uses a deterministic RNG seeded from the given value.
60pub(super) fn generate_random_projections(nproj: usize, m: usize, seed: Option<u64>) -> Vec<f64> {
61    let mut rng: Box<dyn RngCore> = match seed {
62        Some(s) => Box::new(StdRng::seed_from_u64(s)),
63        None => Box::new(rand::thread_rng()),
64    };
65    let mut projections = vec![0.0; nproj * m];
66    for p_idx in 0..nproj {
67        let base = p_idx * m;
68        let mut norm_sq = 0.0;
69        for t in 0..m {
70            let v: f64 = rng.sample(StandardNormal);
71            projections[base + t] = v;
72            norm_sq += v * v;
73        }
74        let inv_norm = 1.0 / norm_sq.sqrt();
75        for t in 0..m {
76            projections[base + t] *= inv_norm;
77        }
78    }
79    projections
80}
81
82/// Project each reference curve onto each projection direction and sort.
83///
84/// Returns a flat buffer of length `nproj * nori` where the sorted projections
85/// for direction `p` occupy `[p*nori .. (p+1)*nori]`.
86pub(super) fn project_and_sort_reference(
87    data_ori: &FdMatrix,
88    projections: &[f64],
89    nproj: usize,
90    nori: usize,
91    m: usize,
92) -> Vec<f64> {
93    let mut sorted = vec![0.0; nproj * nori];
94    maybe_par_chunks_mut_enumerate!(sorted, nori, |(p_idx, spo): (usize, &mut [f64])| {
95        let proj = &projections[p_idx * m..(p_idx + 1) * m];
96        for j in 0..nori {
97            let mut dot = 0.0;
98            for t in 0..m {
99                dot += data_ori[(j, t)] * proj[t];
100            }
101            spo[j] = dot;
102        }
103        spo.sort_unstable_by(|a, b| a.partial_cmp(b).unwrap_or(std::cmp::Ordering::Equal));
104    });
105    sorted
106}
107
108/// Shared implementation for random projection-based depth measures.
109///
110/// Generates `nproj` random projections, projects both object and reference
111/// curves to scalars, and computes univariate depth via binary-search ranking.
112/// The `aggregate` and `finalize` closures control how per-projection depths
113/// are combined (e.g. average for RP depth, minimum for Tukey depth).
114pub(super) fn random_depth_core(
115    data_obj: &FdMatrix,
116    data_ori: &FdMatrix,
117    nproj: usize,
118    seed: Option<u64>,
119    init: f64,
120    aggregate: impl Fn(f64, f64) -> f64 + Sync,
121    finalize: impl Fn(f64, usize) -> f64 + Sync,
122) -> Vec<f64> {
123    use crate::iter_maybe_parallel;
124    #[cfg(feature = "parallel")]
125    use rayon::iter::ParallelIterator;
126
127    let nobj = data_obj.nrows();
128    let nori = data_ori.nrows();
129    let m = data_obj.ncols();
130
131    if nobj == 0 || nori == 0 || m == 0 || nproj == 0 {
132        return Vec::new();
133    }
134
135    let projections = generate_random_projections(nproj, m, seed);
136    let sorted_proj_ori = project_and_sort_reference(data_ori, &projections, nproj, nori, m);
137    let denom = nori as f64 + 1.0;
138
139    iter_maybe_parallel!(0..nobj)
140        .map(|i| {
141            let mut acc = init;
142            for p_idx in 0..nproj {
143                let proj = &projections[p_idx * m..(p_idx + 1) * m];
144                let sorted_ori = &sorted_proj_ori[p_idx * nori..(p_idx + 1) * nori];
145
146                let mut proj_i = 0.0;
147                for t in 0..m {
148                    proj_i += data_obj[(i, t)] * proj[t];
149                }
150
151                let below = sorted_ori.partition_point(|&v| v < proj_i);
152                let above = nori - sorted_ori.partition_point(|&v| v <= proj_i);
153                let depth = (below.min(above) as f64 + 1.0) / denom;
154                acc = aggregate(acc, depth);
155            }
156            finalize(acc, nproj)
157        })
158        .collect()
159}
fdars_core/depth/mod.rs

fdars_core/depth/
mod.rs
