brumby 0.7.3

Derivatives and multiples pricing for racing and sports.
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

use std::ops::RangeInclusive;

use crate::capture::Capture;
use crate::comb::{count_permutations, pick};

#[derive(Clone, Debug)]
pub struct UnivariateDescentConfig {
    pub init_value: f64,
    pub init_step: f64,
    pub min_step: f64,
    pub max_steps: u64,
    pub acceptable_residual: f64,
}

#[derive(Debug)]
pub struct UnivariateDescentOutcome {
    pub steps: u64,
    pub optimal_value: f64,
    pub optimal_residual: f64,
}

/// Univariate, derivative-free search.
pub fn univariate_descent(
    config: &UnivariateDescentConfig,
    mut loss_f: impl FnMut(f64) -> f64,
) -> UnivariateDescentOutcome {
    let mut steps = 0;
    let mut residual = loss_f(config.init_value);
    if residual <= config.acceptable_residual {
        return UnivariateDescentOutcome {
            steps: 0,
            optimal_value: config.init_value,
            optimal_residual: residual
        };
    }

    let (mut value, mut step) = (config.init_value, config.init_step);
    // println!("initial value: {value}, residual: {residual}, step: {step}");
    let (mut optimal_value, mut optimal_residual) = (value, residual);
    // let mut boost = 1.0;
    // let mut gradient: f64 = 1.0;
    while steps < config.max_steps {
        steps += 1;
        let new_value = value + step;/* * boost*/ // * f64::min(gradient.abs(), 100.0);
        let new_residual = loss_f(new_value);
        // let gradient = (new_residual - residual) / (new_value - value);
        // println!("iterations: {iterations}, value: {value}, residual: {residual}, step: {step}, new_value: {new_value}, new_residual: {new_residual}");

        if new_residual > residual {
            step = -step * 0.5;
            if step.abs() < config.min_step {
                break;
            }
        } else if new_residual < optimal_residual {
            // boost = f64::min(gradient.abs(), 10.0);
            // println!("optimal_residual: {optimal_residual}, new_residual: {new_residual}, boost: {boost}, diff: {}", optimal_residual - new_residual);
            optimal_residual = new_residual;
            optimal_value = new_value;

            if optimal_residual <= config.acceptable_residual {
                break;
            }
        }
        residual = new_residual;
        value = new_value;
    }
    UnivariateDescentOutcome {
        steps,
        optimal_value,
        optimal_residual,
    }
}

#[derive(Clone, Debug)]
pub struct HypergridSearchConfig<'a> {
    pub max_steps: u64,
    pub acceptable_residual: f64,
    pub bounds: Capture<'a, Vec<RangeInclusive<f64>>, [RangeInclusive<f64>]>,
    pub resolution: usize,
}

#[derive(Debug)]
pub struct HypergridSearchOutcome {
    pub steps: u64,
    pub optimal_values: Vec<f64>,
    pub optimal_residual: f64,
}

pub fn hypergrid_search(
    config: &HypergridSearchConfig,
    mut constraint_f: impl FnMut(&[f64]) -> bool,
    mut loss_f: impl FnMut(&[f64]) -> f64) -> HypergridSearchOutcome {

    let mut steps = 0;
    let mut values = Vec::with_capacity(config.bounds.len());
    values.resize(values.capacity(), 0.0);

    let mut optimal_values = values.clone();
    let mut optimal_residual = f64::MAX;

    let cardinalities = {
        let mut cardinalities = Vec::with_capacity(values.len());
        cardinalities.resize(cardinalities.capacity(), config.resolution);
        cardinalities
    };
    let mut ordinals = cardinalities.clone();
    let permutations = count_permutations(&cardinalities);
    let mut bounds = (*config.bounds).to_vec();
    let inv_resolution = 1.0 / (config.resolution - 1) as f64;

    'outer: while steps < config.max_steps {
        // println!("step: {steps}, bounds: {bounds:?}");
        steps += 1;

        for permutation in 0.. permutations {
            pick(&cardinalities, permutation, &mut ordinals);

            for (dimension, &ordinal) in ordinals.iter().enumerate() {
                let bound = &bounds[dimension];
                let range = bound.end() - bound.start();
                values[dimension] = bound.start() + ordinal as f64 * range * inv_resolution;
                if constraint_f(&values) {
                    let residual = loss_f(&values);
                    // println!("  values: {values:?}, residual: {residual}");
                    if residual < optimal_residual {
                        // println!("    new optimal");
                        optimal_residual = residual;
                        optimal_values.copy_from_slice(&values);

                        if residual <= config.acceptable_residual {
                            break 'outer;
                        }
                    }
                }
            }
        }

        for (dimension, &value) in optimal_values.iter().enumerate() {
            let hard_bound = &config.bounds[dimension];
            let bound = &mut bounds[dimension];
            let new_range = (bound.end() - bound.start()) / config.resolution as f64;
            let new_start = f64::max(*hard_bound.start(), value - new_range / 2.0);
            let new_end = f64::min(new_start + new_range, *hard_bound.end());
            // let new_start = value - new_range / 2.0;
            // let new_end = new_start + new_range;
            *bound = new_start..=new_end;
        }
    }

    HypergridSearchOutcome {
        steps,
        optimal_values,
        optimal_residual,
    }
}

#[cfg(test)]
mod tests;