radiate-gp 1.3.0

Extensions for radiate. Genetic Programming implementations for graphs (neural networks) and trees.
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

use super::{DataSet, Loss};
use crate::{Graph, GraphChromosome, GraphEvaluator, Op, Tree, TreeChromosome, eval::EvalIntoMut};
use radiate_core::{BatchFitnessFunction, Genotype, fitness::FitnessFunction};
use std::cell::RefCell;

thread_local! {
    static LOSS_BUFFER: RefCell<Vec<f32>> = const { RefCell::new(Vec::new()) };
}

#[derive(Clone)]
pub struct Regression {
    data_set: DataSet<f32>,
    loss: Loss,
}

impl Regression {
    pub fn new(sample_set: impl Into<DataSet<f32>>, loss: Loss) -> Self {
        Regression {
            data_set: sample_set.into(),
            loss,
        }
    }

    #[inline]
    fn calc_into_buff_mut<EV>(&self, eval: &mut EV) -> f32
    where
        EV: EvalIntoMut<[f32], [f32]>,
    {
        let out_len = self.data_set.shape().2;
        LOSS_BUFFER.with(|cell| {
            let mut buf = cell.borrow_mut();

            if buf.len() < out_len {
                buf.resize(out_len, 0.0);
            }

            self.loss
                .calculate(&self.data_set, &mut buf[..out_len], |x, y| {
                    eval.eval_into_mut(x, y)
                })
        })
    }
}

/// --- Graphs ---
impl<'a> FitnessFunction<&'a Genotype<GraphChromosome<Op<f32>>>, f32> for Regression {
    #[inline]
    fn evaluate(&self, input: &'a Genotype<GraphChromosome<Op<f32>>>) -> f32 {
        let mut evaluator = GraphEvaluator::new(&input[0]);
        self.calc_into_buff_mut(&mut evaluator)
    }
}

impl FitnessFunction<Graph<Op<f32>>, f32> for Regression {
    #[inline]
    fn evaluate(&self, input: Graph<Op<f32>>) -> f32 {
        let mut evaluator = GraphEvaluator::new(&input);
        self.calc_into_buff_mut(&mut evaluator)
    }
}

impl BatchFitnessFunction<Graph<Op<f32>>, f32> for Regression {
    #[inline]
    fn evaluate(&self, inputs: Vec<Graph<Op<f32>>>) -> Vec<f32> {
        let mut results = Vec::with_capacity(inputs.len());
        for input in inputs {
            let mut evaluator = GraphEvaluator::new(&input);
            results.push(self.calc_into_buff_mut(&mut evaluator));
        }

        results
    }
}

impl<'a> BatchFitnessFunction<&'a Genotype<GraphChromosome<Op<f32>>>, f32> for Regression {
    #[inline]
    fn evaluate(&self, inputs: Vec<&'a Genotype<GraphChromosome<Op<f32>>>>) -> Vec<f32> {
        let mut results = Vec::with_capacity(inputs.len());
        for input in inputs {
            let mut evaluator = GraphEvaluator::new(&input[0]);
            results.push(self.calc_into_buff_mut(&mut evaluator));
        }

        results
    }
}

/// --- Trees ---
impl FitnessFunction<Tree<Op<f32>>, f32> for Regression {
    #[inline]
    fn evaluate(&self, mut input: Tree<Op<f32>>) -> f32 {
        self.calc_into_buff_mut(&mut input)
    }
}

impl FitnessFunction<Vec<Tree<Op<f32>>>, f32> for Regression {
    #[inline]
    fn evaluate(&self, mut input: Vec<Tree<Op<f32>>>) -> f32 {
        self.calc_into_buff_mut(&mut input)
    }
}

impl BatchFitnessFunction<Tree<Op<f32>>, f32> for Regression {
    #[inline]
    fn evaluate(&self, mut inputs: Vec<Tree<Op<f32>>>) -> Vec<f32> {
        let mut results = Vec::with_capacity(inputs.len());
        for input in inputs.iter_mut() {
            results.push(self.calc_into_buff_mut(input));
        }

        results
    }
}

impl BatchFitnessFunction<Vec<Tree<Op<f32>>>, f32> for Regression {
    #[inline]
    fn evaluate(&self, mut inputs: Vec<Vec<Tree<Op<f32>>>>) -> Vec<f32> {
        let mut results = Vec::with_capacity(inputs.len());
        for input in inputs.iter_mut() {
            results.push(self.calc_into_buff_mut(input));
        }

        results
    }
}

impl<'a> FitnessFunction<&'a Genotype<TreeChromosome<Op<f32>>>, f32> for Regression {
    #[inline]
    fn evaluate(&self, input: &'a Genotype<TreeChromosome<Op<f32>>>) -> f32 {
        let roots = input.iter().map(|c| c.root()).collect::<Vec<_>>();
        self.calc_into_buff_mut(&mut roots.as_slice())
    }
}

impl<'a> BatchFitnessFunction<&'a Genotype<TreeChromosome<Op<f32>>>, f32> for Regression {
    #[inline]
    fn evaluate(&self, inputs: Vec<&'a Genotype<TreeChromosome<Op<f32>>>>) -> Vec<f32> {
        let mut results = Vec::with_capacity(inputs.len());
        for input in inputs {
            let roots = input.iter().map(|c| c.root()).collect::<Vec<_>>();
            results.push(self.calc_into_buff_mut(&mut roots.as_slice()));
        }

        results
    }
}