Line data Source code
1 : // ***************************************************************
2 : // SPDX-FileCopyrightText: Copyright 2024 Ricardo Montañana Gómez
3 : // SPDX-FileType: SOURCE
4 : // SPDX-License-Identifier: MIT
5 : // ***************************************************************
6 :
7 : #include "Node.h"
8 :
9 : namespace bayesnet {
10 :
11 31339 : Node::Node(const std::string& name)
12 31339 : : name(name)
13 : {
14 31339 : }
15 9 : void Node::clear()
16 : {
17 9 : parents.clear();
18 9 : children.clear();
19 9 : cpTable = torch::Tensor();
20 9 : dimensions.clear();
21 9 : numStates = 0;
22 9 : }
23 150429643 : std::string Node::getName() const
24 : {
25 150429643 : return name;
26 : }
27 59262 : void Node::addParent(Node* parent)
28 : {
29 59262 : parents.push_back(parent);
30 59262 : }
31 17 : void Node::removeParent(Node* parent)
32 : {
33 17 : parents.erase(std::remove(parents.begin(), parents.end(), parent), parents.end());
34 17 : }
35 17 : void Node::removeChild(Node* child)
36 : {
37 17 : children.erase(std::remove(children.begin(), children.end(), child), children.end());
38 17 : }
39 59235 : void Node::addChild(Node* child)
40 : {
41 59235 : children.push_back(child);
42 59235 : }
43 5087 : std::vector<Node*>& Node::getParents()
44 : {
45 5087 : return parents;
46 : }
47 77571 : std::vector<Node*>& Node::getChildren()
48 : {
49 77571 : return children;
50 : }
51 64124 : int Node::getNumStates() const
52 : {
53 64124 : return numStates;
54 : }
55 32864 : void Node::setNumStates(int numStates)
56 : {
57 32864 : this->numStates = numStates;
58 32864 : }
59 429 : torch::Tensor& Node::getCPT()
60 : {
61 429 : return cpTable;
62 : }
63 : /*
64 : The MinFill criterion is a heuristic for variable elimination.
65 : The variable that minimizes the number of edges that need to be added to the graph to make it triangulated.
66 : This is done by counting the number of edges that need to be added to the graph if the variable is eliminated.
67 : The variable with the minimum number of edges is chosen.
68 : Here this is done computing the length of the combinations of the node neighbors taken 2 by 2.
69 : */
70 45 : unsigned Node::minFill()
71 : {
72 45 : std::unordered_set<std::string> neighbors;
73 117 : for (auto child : children) {
74 72 : neighbors.emplace(child->getName());
75 : }
76 108 : for (auto parent : parents) {
77 63 : neighbors.emplace(parent->getName());
78 : }
79 45 : auto source = std::vector<std::string>(neighbors.begin(), neighbors.end());
80 90 : return combinations(source).size();
81 45 : }
82 45 : std::vector<std::pair<std::string, std::string>> Node::combinations(const std::vector<std::string>& source)
83 : {
84 45 : std::vector<std::pair<std::string, std::string>> result;
85 180 : for (int i = 0; i < source.size(); ++i) {
86 135 : std::string temp = source[i];
87 279 : for (int j = i + 1; j < source.size(); ++j) {
88 144 : result.push_back({ temp, source[j] });
89 : }
90 135 : }
91 90 : return result;
92 45 : }
93 32894 : void Node::computeCPT(const torch::Tensor& dataset, const std::vector<std::string>& features, const double laplaceSmoothing, const torch::Tensor& weights)
94 : {
95 32894 : dimensions.clear();
96 : // Get dimensions of the CPT
97 32894 : dimensions.push_back(numStates);
98 94914 : transform(parents.begin(), parents.end(), back_inserter(dimensions), [](const auto& parent) { return parent->getNumStates(); });
99 : // Create a tensor of zeros with the dimensions of the CPT
100 32894 : cpTable = torch::zeros(dimensions, torch::kFloat) + laplaceSmoothing;
101 : // Fill table with counts
102 32894 : auto pos = find(features.begin(), features.end(), name);
103 32894 : if (pos == features.end()) {
104 8 : throw std::logic_error("Feature " + name + " not found in dataset");
105 : }
106 32886 : int name_index = pos - features.begin();
107 11221522 : for (int n_sample = 0; n_sample < dataset.size(1); ++n_sample) {
108 11188649 : c10::List<c10::optional<at::Tensor>> coordinates;
109 33565947 : coordinates.push_back(dataset.index({ name_index, n_sample }));
110 32200749 : for (auto parent : parents) {
111 21012113 : pos = find(features.begin(), features.end(), parent->getName());
112 21012113 : if (pos == features.end()) {
113 13 : throw std::logic_error("Feature parent " + parent->getName() + " not found in dataset");
114 : }
115 21012100 : int parent_index = pos - features.begin();
116 63036300 : coordinates.push_back(dataset.index({ parent_index, n_sample }));
117 : }
118 : // Increment the count of the corresponding coordinate
119 22377272 : cpTable.index_put_({ coordinates }, cpTable.index({ coordinates }) + weights.index({ n_sample }).item<double>());
120 11188649 : }
121 : // Normalize the counts
122 32873 : cpTable = cpTable / cpTable.sum(0);
123 43422258 : }
124 69151761 : float Node::getFactorValue(std::map<std::string, int>& evidence)
125 : {
126 69151761 : c10::List<c10::optional<at::Tensor>> coordinates;
127 : // following predetermined order of indices in the cpTable (see Node.h)
128 69151761 : coordinates.push_back(at::tensor(evidence[name]));
129 198453273 : transform(parents.begin(), parents.end(), std::back_inserter(coordinates), [&evidence](const auto& parent) { return at::tensor(evidence[parent->getName()]); });
130 138303522 : return cpTable.index({ coordinates }).item<float>();
131 69151761 : }
132 732 : std::vector<std::string> Node::graph(const std::string& className)
133 : {
134 732 : auto output = std::vector<std::string>();
135 732 : auto suffix = name == className ? ", fontcolor=red, fillcolor=lightblue, style=filled " : "";
136 732 : output.push_back(name + " [shape=circle" + suffix + "] \n");
137 1840 : transform(children.begin(), children.end(), back_inserter(output), [this](const auto& child) { return name + " -> " + child->getName(); });
138 1464 : return output;
139 732 : }
140 : }
|
