dogs 1.3.0

Discrete Optimization Global Search framework. Implements various search algorithms that can be found in combinatorial optimization or heuristic search.
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

use min_max_heap::MinMaxHeap;
use std::cmp::{Ord, PartialOrd};
use std::marker::PhantomData;
use std::cell::RefCell;
use std::rc::Rc;

use crate::search_manager::SearchManager;
use crate::search_algorithm::{StoppingCriterion, SearchAlgorithm};
use crate::tree_search::algo::helper::guided_node::GuidedNode;
use crate::search_space::{
    SearchSpace, GuidedSpace, TotalNeighborGeneration, ParetoDominanceSpace
};

/**
 * implements beam dominance schemes. i.e. the policy to keep track of "elite" nodes that will be
 * used to eliminate dominated nodes.
 */
#[derive(Debug, Clone)]
pub enum BeamDominanceScheme {
    /// keep a maximum number of elite nodes
    Fixed(usize)
}

/** beam search with pareto-dominance scheme */
#[derive(Debug)]
pub struct BeamSearchDom<N, B, G, Space> {
    manager: SearchManager<N, B>,
    space: Rc<RefCell<Space>>,
    d: usize,
    heuristic_pruning_done: bool,
    beam_dominance_scheme: BeamDominanceScheme,
    g: PhantomData<G>,
}

impl<Space, N:Clone, B:PartialOrd+Copy, G:Ord> BeamSearchDom<N, B, G, Space> {
    /** builds the beam search using the search space, the beam width, and the dominance scheme */
    pub fn new(space: Rc<RefCell<Space>>, d: usize, dom_scheme:BeamDominanceScheme) -> Self {
        Self {
            manager: SearchManager::default(),
            space,
            d,
            heuristic_pruning_done: false,
            beam_dominance_scheme: dom_scheme,
            g: PhantomData,
        }
    }
}

impl<'a, N, B, G, Space> SearchAlgorithm<N, B> for BeamSearchDom<N, B, G, Space>
where
    N: Clone,
    B: PartialOrd+Copy,
    G: Ord+Clone,
    Space: SearchSpace<N,B> + GuidedSpace<N,G> + TotalNeighborGeneration<N> + ParetoDominanceSpace<N>,
{
    fn run<SC:StoppingCriterion>(&mut self, stopping_criterion:SC) {
        let mut space = self.space.borrow_mut();
        let mut beam = MinMaxHeap::with_capacity(self.d);
        let root = space.initial();
        let g_root = space.guide(&root);
        self.heuristic_pruning_done = false;
        beam.push(GuidedNode::new(root, g_root));
        while !stopping_criterion.is_finished() && !beam.is_empty() {
            let mut next_beam = MinMaxHeap::with_capacity(self.d);
            let mut elites:MinMaxHeap<GuidedNode<N,G>> = MinMaxHeap::new();
            while !beam.is_empty() && !stopping_criterion.is_finished() {
                let mut n = beam.pop_min().unwrap().node;
                // check if goal
                if space.goal(&n) {
                    // compare with best
                    let v = space.bound(&n);
                    if self.manager.is_better(v) {
                        let n2 = space.handle_new_best(n);
                        n = n2.clone();
                        let b2 = space.bound(&n2);
                        self.manager.update_best(n2, b2);
                    }
                }
                let mut children = space.neighbors(&mut n);
                while !children.is_empty() {
                    let c = children.pop().unwrap();
                    // check if goal
                    if space.goal(&c) {
                        // compare with best
                        let v = space.bound(&c);
                        if self.manager.is_better(v) {
                            let c2 = space.handle_new_best(c);
                            let b2 = space.bound(&c2);
                            self.manager.update_best(c2, b2);
                        }
                        continue;
                    }
                    // check if n is dominated by the elite set
                    let mut is_dominated = false;
                    for e in elites.iter() {
                        if space.dominates(&e.node, &c) {
                            is_dominated = true;
                            continue;
                        }
                    }
                    if !is_dominated {
                        let c_guide = space.guide(&c);
                        // possibly: add n to the elite set
                        let BeamDominanceScheme::Fixed(elite_max_size) = self.beam_dominance_scheme;
                        if elites.len() < elite_max_size {
                            elites.push(GuidedNode::new(c.clone(), c_guide.clone()));
                        } else {
                            elites.push_pop_max(GuidedNode::new(c.clone(), c_guide.clone()));
                        }
                         // compute guide to feed the GuidedNode while inserting into next_beam
                        if next_beam.len() < self.d {
                            next_beam.push(GuidedNode::new(c, c_guide));
                        } else {
                            self.heuristic_pruning_done = true;
                            // pop max and insert child
                            next_beam.push_pop_max(GuidedNode::new(c, c_guide));
                        }
                    }
                }
            }
            beam = next_beam;
        }
        space.stop_search("".to_string());
    }

    fn get_manager(&mut self) -> &mut SearchManager<N, B> { &mut self.manager }

    /**
     * returns true if the optimal value is found (thus we can stop the search)
     */
    fn is_optimal(&self) -> bool { !self.heuristic_pruning_done }
}