optimization-solvers 0.1.1

Numerical optimization solvers for unconstrained and simple-bounds constrained convex optimization problems. Wasm compatible
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

use super::*;

pub trait ComputeDirection {
    fn compute_direction(
        &mut self,
        eval_x_k: &FuncEvalMultivariate,
    ) -> Result<DVector<Floating>, SolverError>;
}

#[derive(thiserror::Error, Debug)]
pub enum SolverError {
    #[error("Max iter reached")]
    MaxIterReached,
    #[error("Out of domain")]
    OutOfDomain,
    #[error("Error in input parameters")]
    ErrorInputParams,
    #[error("Abnormal termination")]
    AbnormalTermination,
}

//Template pattern for solvers. Methods that are already implemented can be freely overriden.
pub trait LineSearchSolver: ComputeDirection {
    fn xk(&self) -> &DVector<Floating>;
    fn xk_mut(&mut self) -> &mut DVector<Floating>;
    fn k(&self) -> &usize;
    fn k_mut(&mut self) -> &mut usize;
    fn has_converged(&self, eval_x_k: &FuncEvalMultivariate) -> bool;

    fn setup(&mut self) {}

    fn evaluate_x_k(
        &mut self,
        oracle: &mut impl FnMut(&DVector<Floating>) -> FuncEvalMultivariate,
    ) -> Result<FuncEvalMultivariate, SolverError> {
        let eval_x_k = oracle(self.xk());
        if eval_x_k.f().is_nan() || eval_x_k.f().is_infinite() {
            error!(target: "solver","Minimization completed: next iterate is out of domain");
            return Err(SolverError::OutOfDomain);
        }
        Ok(eval_x_k)
    }

    fn update_next_iterate<LS: LineSearch>(
        &mut self,
        line_search: &mut LS,
        eval_x_k: &FuncEvalMultivariate, //eval_x_k: &FuncEvalMultivariate,
        oracle: &mut impl FnMut(&DVector<Floating>) -> FuncEvalMultivariate,
        direction: &DVector<Floating>,
        max_iter_line_search: usize,
    ) -> Result<(), SolverError> {
        let step = line_search.compute_step_len(
            self.xk(),
            eval_x_k,
            direction,
            oracle,
            max_iter_line_search,
        );

        let next_iterate = self.xk() + step * direction;
        *self.xk_mut() = next_iterate;

        Ok(())
    }

    fn minimize<LS: LineSearch>(
        &mut self,
        line_search: &mut LS,
        mut oracle: impl FnMut(&DVector<Floating>) -> FuncEvalMultivariate,
        max_iter_solver: usize,
        max_iter_line_search: usize,
        mut callback: Option<&mut dyn FnMut(&Self)>,
    ) -> Result<(), SolverError> {
        *self.k_mut() = 0;

        self.setup();

        while &max_iter_solver > self.k() {
            let eval_x_k = self.evaluate_x_k(&mut oracle)?;

            if self.has_converged(&eval_x_k) {
                info!(
                    target: "solver",
                    "Minimization completed: convergence in {} iterations",
                    self.k()
                );
                return Ok(());
            }

            let direction = self.compute_direction(&eval_x_k)?;

            debug!(target: "solver","Gradient: {:?}, Direction: {:?}", eval_x_k.g(), direction);
            self.update_next_iterate(
                line_search,
                &eval_x_k,
                &mut oracle,
                &direction,
                max_iter_line_search,
            )?;

            debug!(target: "solver","Iterate: {:?}", self.xk());
            debug!(target: "solver","Function eval: {:?}", eval_x_k);

            *self.k_mut() += 1;
            if let Some(callback) = callback.as_mut() {
                callback(self);
            }
        }
        warn!(target: "solver","Minimization completed: max iter reached during minimization");
        Err(SolverError::MaxIterReached)
    }
}

pub trait HasBounds {
    fn lower_bound(&self) -> &DVector<Floating>;
    fn upper_bound(&self) -> &DVector<Floating>;
    fn set_lower_bound(&mut self, lower_bound: DVector<Floating>);
    fn set_upper_bound(&mut self, upper_bound: DVector<Floating>);
}

pub trait HasProjectedGradient: LineSearchSolver + HasBounds {
    fn projected_gradient(&self, eval: &FuncEvalMultivariate) -> DVector<Floating> {
        let mut proj_grad = eval.g().clone();
        for (i, x) in self.xk().iter().enumerate() {
            if (x == &self.lower_bound()[i] && proj_grad[i] > 0.0)
                || (x == &self.upper_bound()[i] && proj_grad[i] < 0.0)
            {
                proj_grad[i] = 0.0;
            }
        }
        proj_grad
    }
}

//Blanket implementation for all optimization solvers that have bounds
impl<T> HasProjectedGradient for T where T: LineSearchSolver + HasBounds {}