givp 1.0.0

GRASP-ILS-VND with Path Relinking metaheuristic for continuous black-box optimization
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
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197

// SPDX-FileCopyrightText: 2026 Arnaldo Mendes Pires Junior
// SPDX-License-Identifier: MIT

use crate::cache::EvaluationCache;
use crate::grasp::evaluate_with_cache;
use crate::helpers::expired;
use rand::Rng;
use rand_chacha::ChaCha8Rng;
use std::time::Instant;

const MAX_PR_VARS: usize = 25;

/// Best (greedy) path relinking: at each step pick the variable move with best cost.
fn path_relinking_best<F>(
    func: &F,
    source: &[f64],
    target: &[f64],
    diff_indices: &[usize],
    cache: &mut Option<EvaluationCache>,
    half: usize,
    deadline: Option<Instant>,
) -> (Vec<f64>, f64)
where
    F: Fn(&[f64]) -> f64,
{
    let mut current = source.to_vec();
    let mut best = current.clone();
    let mut best_cost = evaluate_with_cache(&current, func, cache, half);
    let mut remaining: Vec<usize> = diff_indices.to_vec();

    while !remaining.is_empty() {
        if expired(deadline) {
            break;
        }
        let mut best_idx_pos = 0;
        let mut best_move_cost = f64::INFINITY;
        let mut best_move_val = 0.0;

        for (pos, &idx) in remaining.iter().enumerate() {
            let old = current[idx];
            current[idx] = target[idx];
            let cost = evaluate_with_cache(&current, func, cache, half);
            if cost < best_move_cost {
                best_move_cost = cost;
                best_idx_pos = pos;
                best_move_val = target[idx];
            }
            current[idx] = old;
        }

        let chosen_idx = remaining.swap_remove(best_idx_pos);
        current[chosen_idx] = best_move_val;

        if best_move_cost < best_cost {
            best_cost = best_move_cost;
            best = current.clone();
        }
    }
    (best, best_cost)
}

/// Bidirectional path relinking between two solutions.
pub(crate) fn bidirectional_path_relinking<F>(
    func: &F,
    sol1: &[f64],
    sol2: &[f64],
    half: usize,
    cache: &mut Option<EvaluationCache>,
    rng: &mut ChaCha8Rng,
    deadline: Option<Instant>,
) -> (Vec<f64>, f64)
where
    F: Fn(&[f64]) -> f64,
{
    let n = sol1.len();
    let mut diff_indices: Vec<usize> = (0..n)
        .filter(|&i| (sol1[i] - sol2[i]).abs() > 1e-12)
        .collect();

    if diff_indices.is_empty() {
        let cost = evaluate_with_cache(sol1, func, cache, half);
        return (sol1.to_vec(), cost);
    }

    // Limit to top MAX_PR_VARS most different variables
    if diff_indices.len() > MAX_PR_VARS {
        diff_indices.sort_by(|&a, &b| {
            let da = (sol1[a] - sol2[a]).abs();
            let db = (sol1[b] - sol2[b]).abs();
            db.partial_cmp(&da).unwrap()
        });
        diff_indices.truncate(MAX_PR_VARS);
    }

    // Shuffle order
    for i in (1..diff_indices.len()).rev() {
        let j = rng.random_range(0..=i);
        diff_indices.swap(i, j);
    }

    // Forward: sol1 → sol2 (best strategy)
    let (best_fwd, cost_fwd) =
        path_relinking_best(func, sol1, sol2, &diff_indices, cache, half, deadline);

    // Backward: sol2 → sol1 (best strategy)
    let (best_bwd, cost_bwd) =
        path_relinking_best(func, sol2, sol1, &diff_indices, cache, half, deadline);

    if cost_fwd <= cost_bwd {
        (best_fwd, cost_fwd)
    } else {
        (best_bwd, cost_bwd)
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use rand::SeedableRng;
    use rand_chacha::ChaCha8Rng;
    use std::time::{Duration, Instant};

    #[test]
    fn test_identical_solutions_returns_immediately() {
        let mut rng = ChaCha8Rng::seed_from_u64(0);
        let sol = vec![1.0, 2.0, 3.0];
        let (result, cost) = bidirectional_path_relinking(
            &|x: &[f64]| x.iter().sum::<f64>(),
            &sol,
            &sol,
            3,
            &mut None,
            &mut rng,
            None,
        );
        assert_eq!(result, sol);
        assert!((cost - 6.0).abs() < 1e-10);
    }

    #[test]
    fn test_expired_deadline_breaks_inner_loop() {
        // deadline already expired → path_relinking_best loop breaks at line 33
        let mut rng = ChaCha8Rng::seed_from_u64(0);
        let sol1 = vec![0.0, 0.0, 0.0];
        let sol2 = vec![1.0, 1.0, 1.0];
        let func = |x: &[f64]| x.iter().map(|&xi| xi * xi).sum::<f64>();
        assert!((func(&sol1) - 0.0).abs() < 1e-10); // invoke closure body
        let deadline = Some(Instant::now() - Duration::from_secs(1));
        let (result, _cost) =
            bidirectional_path_relinking(&func, &sol1, &sol2, 3, &mut None, &mut rng, deadline);
        assert_eq!(result.len(), 3);
    }

    /// Constructs a function on the {0,1}^3 grid where the backward greedy path
    /// finds [0,1,1] (cost 2.0) — a point the forward greedy path never visits.
    /// Forward best = 5.0 (sol2), backward best = 2.0 → backward wins.
    #[test]
    fn test_backward_path_wins() {
        let mut rng = ChaCha8Rng::seed_from_u64(0);
        let func = |x: &[f64]| {
            let a = x[0] >= 0.5;
            let b = x[1] >= 0.5;
            let c = x[2] >= 0.5;
            match (a, b, c) {
                (false, false, false) => 10.0_f64, // sol1
                (true, false, false) => 8.0,
                (false, true, false) => 9.0,
                (false, false, true) => 11.0,
                (true, true, false) => 6.0,
                (true, false, true) => 7.0,
                (false, true, true) => 2.0, // backward step-1 finds this; forward never does
                (true, true, true) => 5.0,  // sol2
            }
        };
        let sol1 = vec![0.0, 0.0, 0.0];
        let sol2 = vec![1.0, 1.0, 1.0];
        let (result, cost) =
            bidirectional_path_relinking(&func, &sol1, &sol2, 3, &mut None, &mut rng, None);
        // Backward path: [1,1,1]→ set x[0]=0 → [0,1,1]=2.0 (best) → backward wins
        assert!((cost - 2.0).abs() < 1e-10);
        assert!((result[0]).abs() < 1e-10);
        assert!((result[1] - 1.0).abs() < 1e-10);
        assert!((result[2] - 1.0).abs() < 1e-10);
    }

    #[test]
    fn test_truncates_when_many_differences() {
        let mut rng = ChaCha8Rng::seed_from_u64(1);
        let sol1 = vec![0.0; 40];
        let sol2 = vec![1.0; 40];
        let func = |x: &[f64]| x.iter().sum::<f64>();
        let (result, cost) =
            bidirectional_path_relinking(&func, &sol1, &sol2, 40, &mut None, &mut rng, None);
        assert_eq!(result.len(), 40);
        assert!(cost.is_finite());
    }
}