resopt 0.3.0

Declarative constrained residual optimization 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

use crate::core::{ConstrainedResidualProblem, Error, Matrix, TikhonovRegularization};

pub fn compute_p_dense(matrix: &Matrix) -> Vec<f64> {
    let rows = matrix.nrows();
    let cols = matrix.ncols();
    let data = matrix.data();

    let mut p = vec![0.0; cols * cols];

    for i in 0..cols {
        for j in 0..cols {
            let mut sum = 0.0;
            for k in 0..rows {
                sum += data[k * cols + i] * data[k * cols + j];
            }
            p[i * cols + j] = sum;
        }
    }

    p
}

pub fn compute_q(matrix: &Matrix, target: &[f64]) -> Vec<f64> {
    let rows = matrix.nrows();
    let cols = matrix.ncols();
    let data = matrix.data();

    let mut q = vec![0.0; cols];

    for j in 0..cols {
        let mut sum = 0.0;
        for i in 0..rows {
            sum += data[i * cols + j] * target[i];
        }
        q[j] = -sum;
    }

    q
}

pub fn compute_residual(matrix: &Matrix, x: &[f64], target: &[f64]) -> Vec<f64> {
    let rows = matrix.nrows();
    let cols = matrix.ncols();
    let data = matrix.data();

    let mut residual = vec![0.0; rows];

    for i in 0..rows {
        let mut value = 0.0;
        for j in 0..cols {
            value += data[i * cols + j] * x[j];
        }
        residual[i] = value - target[i];
    }

    residual
}

pub fn l2_squared_value(residual: &[f64]) -> f64 {
    0.5 * residual.iter().map(|value| value * value).sum::<f64>()
}

pub fn tikhonov_value(regularization: &TikhonovRegularization, x: &[f64]) -> f64 {
    let residual = compute_residual(regularization.matrix(), x, regularization.target());
    regularization.lambda() * l2_squared_value(&residual)
}

pub fn apply_matrix(matrix: &Matrix, x: &[f64]) -> Vec<f64> {
    let rows = matrix.nrows();
    let cols = matrix.ncols();
    let data = matrix.data();

    let mut out = vec![0.0; rows];

    for i in 0..rows {
        let mut value = 0.0;
        for j in 0..cols {
            value += data[i * cols + j] * x[j];
        }
        out[i] = value;
    }

    out
}

pub fn max_constraint_violation(problem: &ConstrainedResidualProblem, x: &[f64]) -> f64 {
    let mut max_violation: f64 = 0.0;

    for eq in problem.equalities() {
        let vals = apply_matrix(eq.matrix(), x);
        for (lhs, rhs) in vals.iter().zip(eq.rhs().iter()) {
            max_violation = max_violation.max((lhs - rhs).abs());
        }
    }

    for ineq in problem.inequalities() {
        let vals = apply_matrix(ineq.matrix(), x);
        for (lhs, rhs) in vals.iter().zip(ineq.rhs().iter()) {
            max_violation = max_violation.max((lhs - rhs).max(0.0));
        }
    }

    if let Some(bounds) = problem.bounds() {
        for (i, x_i) in x.iter().copied().enumerate().take(bounds.len()) {
            if let Some(lb) = bounds.lower()[i] {
                max_violation = max_violation.max((lb - x_i).max(0.0));
            }
            if let Some(ub) = bounds.upper()[i] {
                max_violation = max_violation.max((x_i - ub).max(0.0));
            }
        }
    }

    max_violation
}

pub fn copy_block_rows(
    block_matrix: &Matrix,
    a_dense: &mut [f64],
    total_ncols: usize,
    start_row: usize,
) -> Result<(), Error> {
    if block_matrix.ncols() != total_ncols {
        return Err(Error::DimensionMismatch {
            message: format!(
                "constraint block has {} columns but expected {}",
                block_matrix.ncols(),
                total_ncols
            ),
        });
    }

    let rows = block_matrix.nrows();
    let cols = block_matrix.ncols();
    let src = block_matrix.data();

    for i in 0..rows {
        for j in 0..cols {
            a_dense[(start_row + i) * total_ncols + j] = src[i * cols + j];
        }
    }

    Ok(())
}