-
Notifications
You must be signed in to change notification settings - Fork 2
/
example_network.cpp
107 lines (91 loc) · 3.73 KB
/
example_network.cpp
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
//
// Created by paolo on 31/08/21.
//
#include <baylib/network/bayesian_net.hpp>
#include <baylib/inference/rejection_sampling.hpp>
#include <iostream>
/*
* The Bayesian Network Class is the main container of the baylib library.
* This data structure holds both the graph structure and the CPT of each variable inside the network.
* CPT memory usage is optimized with the integration of the Copy On Write paradigm (COW),
* each time a new CPT is added if a duplicate is found the new one is discarded and a reference to the original
* one is saved instead.
*/
int main(){
using namespace baylib;
bayesian_net<random_variable<double>> bn;
// We want to create manually the following
// Bayesain network
// B C
// \ /
// \ /
// v
// A
// |
// |
// v
// D
// Use add_variable to add a new random variable with
// its random states (default 2 states)
ulong A = bn.add_variable(2); // 2 can be omitted
ulong B = bn.add_variable();
ulong C = bn.add_variable();
ulong D = bn.add_variable(4); // D has 4 states
// Use add_dependency to add an edge between two variables
bn.add_dependency(B, A);
bn.add_dependency(C, A);
bn.add_dependency(A, D);
// Condition object is used to specify elements in the cpt
baylib::condition c;
bn.set_variable_probability(B, 0, c, .5);
bn.set_variable_probability(B, 1, c, .5);
bn.set_variable_probability(C, 0, c, .002);
bn.set_variable_probability(C, 1, c, 1 - .002);
// An alternative to building the condition by hand is using the condition factory util to generate
// all the possible conditions of a cpt
std::vector<float> pb_vector = {0.1, 0.9, 0.2, 0.8, 0.3, 0.7, 0.5, 0.5};
auto factory = baylib::condition_factory(bn, A);
uint i = 0;
do {
bn.set_variable_probability(A, 0, factory.get(), pb_vector[i++]);
bn.set_variable_probability(A, 1, factory.get(), pb_vector[i++]);
} while(factory.has_next());
c.add(A, 0);
bn.set_variable_probability(D, 0, c, .5);
bn.set_variable_probability(D, 1, c, .5);
c.add(A, 0);
bn.set_variable_probability(D, 0, c, .9);
bn.set_variable_probability(D, 1, c, .1);
c.add(A, 1);
bn.set_variable_probability(D, 0, c, .9);
bn.set_variable_probability(D, 1, c, .1);
// Bayesian network can be iterated over with a for loop
for (auto &var: bn) {
std::cout << var.id() << '\n';
std::cout << var.table() << '\n';
}
// we can now perform inference on the network
// let's use the rejection sampling approximate inference algorithm
// with 1e5 samples and 10 threads.
// The usage of other algorithms is show in example_inference
// with named_bayesian networks
using namespace baylib::inference;
rejection_sampling rs(bn, 10000, 10);
// the make_inference() method retrieves a marginal distribution
// for the entire network
auto inf_result = rs.make_inference();
// we can now print it as it is using the operator <<
// or prettify it a little
std::cout << "P(A=0) = " << inf_result[A][0] << '\n';
std::cout << "P(A=1) = " << inf_result[A][1] << '\n';
std::cout << "P(B=0) = " << inf_result[B][0] << '\n';
std::cout << "P(B=1) = " << inf_result[B][1] << '\n';
std::cout << "P(C=0) = " << inf_result[C][0] << '\n';
std::cout << "P(C=1) = " << inf_result[C][1] << '\n';
std::cout << "P(D=0) = " << inf_result[D][0] << '\n';
std::cout << "P(D=1) = " << inf_result[D][1] << '\n';
std::cout << "P(D=2) = " << inf_result[D][2] << '\n';
std::cout << "P(D=3) = " << inf_result[D][3] << '\n';
// to see how to add an evidence and make inference
// using that constraint, have a look at example_inference.cpp
}