| 1 | /* ---------------------------------------------------------------------------- |
| 2 | |
| 3 | * GTSAM Copyright 2010, Georgia Tech Research Corporation, |
| 4 | * Atlanta, Georgia 30332-0415 |
| 5 | * All Rights Reserved |
| 6 | * Authors: Frank Dellaert, et al. (see THANKS for the full author list) |
| 7 | |
| 8 | * See LICENSE for the license information |
| 9 | |
| 10 | * -------------------------------------------------------------------------- */ |
| 11 | |
| 12 | /** |
| 13 | * @file DiscreteBayesNet.h |
| 14 | * @date Feb 15, 2011 |
| 15 | * @author Duy-Nguyen Ta |
| 16 | * @author Frank dellaert |
| 17 | */ |
| 18 | |
| 19 | #pragma once |
| 20 | |
| 21 | #include <gtsam/discrete/DiscreteConditional.h> |
| 22 | #include <gtsam/discrete/DiscreteDistribution.h> |
| 23 | #include <gtsam/inference/BayesNet.h> |
| 24 | #include <gtsam/inference/FactorGraph.h> |
| 25 | |
| 26 | #include <memory> |
| 27 | #include <map> |
| 28 | #include <string> |
| 29 | #include <utility> |
| 30 | #include <vector> |
| 31 | |
| 32 | namespace gtsam { |
| 33 | |
| 34 | /** |
| 35 | * A Bayes net made from discrete conditional distributions. |
| 36 | * @ingroup discrete |
| 37 | */ |
| 38 | class GTSAM_EXPORT DiscreteBayesNet: public BayesNet<DiscreteConditional> { |
| 39 | public: |
| 40 | typedef BayesNet<DiscreteConditional> Base; |
| 41 | typedef DiscreteBayesNet This; |
| 42 | typedef DiscreteConditional ConditionalType; |
| 43 | typedef std::shared_ptr<This> shared_ptr; |
| 44 | typedef std::shared_ptr<ConditionalType> sharedConditional; |
| 45 | |
| 46 | /// @name Standard Constructors |
| 47 | /// @{ |
| 48 | |
| 49 | /// Construct empty Bayes net. |
| 50 | DiscreteBayesNet() {} |
| 51 | |
| 52 | /** Construct from iterator over conditionals */ |
| 53 | template <typename ITERATOR> |
| 54 | DiscreteBayesNet(ITERATOR firstConditional, ITERATOR lastConditional) |
| 55 | : Base(firstConditional, lastConditional) {} |
| 56 | |
| 57 | /** Construct from container of factors (shared_ptr or plain objects) */ |
| 58 | template <class CONTAINER> |
| 59 | explicit DiscreteBayesNet(const CONTAINER& conditionals) |
| 60 | : Base(conditionals) {} |
| 61 | |
| 62 | /** Implicit copy/downcast constructor to override explicit template |
| 63 | * container constructor */ |
| 64 | template <class DERIVEDCONDITIONAL> |
| 65 | DiscreteBayesNet(const FactorGraph<DERIVEDCONDITIONAL>& graph) |
| 66 | : Base(graph) {} |
| 67 | |
| 68 | /// @} |
| 69 | |
| 70 | /// @name Testable |
| 71 | /// @{ |
| 72 | |
| 73 | /** Check equality */ |
| 74 | bool equals(const This& bn, double tol = 1e-9) const; |
| 75 | |
| 76 | /// @} |
| 77 | |
| 78 | /// @name Standard Interface |
| 79 | /// @{ |
| 80 | |
| 81 | // Add inherited versions of add. |
| 82 | using Base::add; |
| 83 | |
| 84 | /** Add a DiscreteDistribution using a table or a string */ |
| 85 | void add(const DiscreteKey& key, const std::string& spec) { |
| 86 | emplace_shared<DiscreteDistribution>(args: key, args: spec); |
| 87 | } |
| 88 | |
| 89 | /** Add a DiscreteCondtional */ |
| 90 | template <typename... Args> |
| 91 | void add(Args&&... args) { |
| 92 | emplace_shared<DiscreteConditional>(std::forward<Args>(args)...); |
| 93 | } |
| 94 | |
| 95 | //** evaluate for given DiscreteValues */ |
| 96 | double evaluate(const DiscreteValues & values) const; |
| 97 | |
| 98 | //** (Preferred) sugar for the above for given DiscreteValues */ |
| 99 | double operator()(const DiscreteValues & values) const { |
| 100 | return evaluate(values); |
| 101 | } |
| 102 | |
| 103 | //** log(evaluate(values)) for given DiscreteValues */ |
| 104 | double logProbability(const DiscreteValues & values) const; |
| 105 | |
| 106 | /** |
| 107 | * @brief do ancestral sampling |
| 108 | * |
| 109 | * Assumes the Bayes net is reverse topologically sorted, i.e. last |
| 110 | * conditional will be sampled first. If the Bayes net resulted from |
| 111 | * eliminating a factor graph, this is true for the elimination ordering. |
| 112 | * |
| 113 | * @return a sampled value for all variables. |
| 114 | */ |
| 115 | DiscreteValues sample(std::mt19937_64* rng = nullptr) const; |
| 116 | |
| 117 | /** |
| 118 | * @brief do ancestral sampling, given certain variables. |
| 119 | * |
| 120 | * Assumes the Bayes net is reverse topologically sorted *and* that the |
| 121 | * Bayes net does not contain any conditionals for the given values. |
| 122 | * |
| 123 | * @return given values extended with sampled value for all other variables. |
| 124 | */ |
| 125 | DiscreteValues sample(DiscreteValues given, |
| 126 | std::mt19937_64* rng = nullptr) const; |
| 127 | |
| 128 | /** |
| 129 | * @brief Prune the Bayes net |
| 130 | * |
| 131 | * @param maxNrLeaves The maximum number of leaves to keep. |
| 132 | * @param marginalThreshold If given, threshold on marginals to prune variables. |
| 133 | * @param fixedValues If given, return the fixed values removed. |
| 134 | * @return A new DiscreteBayesNet with pruned conditionals. |
| 135 | */ |
| 136 | DiscreteBayesNet prune(size_t maxNrLeaves, |
| 137 | const std::optional<double>& marginalThreshold = {}, |
| 138 | DiscreteValues* fixedValues = nullptr) const; |
| 139 | |
| 140 | /** |
| 141 | * @brief Multiply all conditionals into one big joint conditional |
| 142 | * and return it. |
| 143 | * |
| 144 | * NOTE: possibly quite expensive. |
| 145 | * |
| 146 | * @return DiscreteConditional |
| 147 | */ |
| 148 | DiscreteConditional joint() const; |
| 149 | |
| 150 | ///@} |
| 151 | /// @name Wrapper support |
| 152 | /// @{ |
| 153 | |
| 154 | /// Render as markdown tables. |
| 155 | std::string markdown(const KeyFormatter& keyFormatter = DefaultKeyFormatter, |
| 156 | const DiscreteFactor::Names& names = {}) const; |
| 157 | |
| 158 | /// Render as html tables. |
| 159 | std::string html(const KeyFormatter& keyFormatter = DefaultKeyFormatter, |
| 160 | const DiscreteFactor::Names& names = {}) const; |
| 161 | |
| 162 | /// @} |
| 163 | /// @name HybridValues methods. |
| 164 | /// @{ |
| 165 | |
| 166 | using Base::error; // Expose error(const HybridValues&) method.. |
| 167 | using Base::evaluate; // Expose evaluate(const HybridValues&) method.. |
| 168 | using Base::logProbability; // Expose logProbability(const HybridValues&) |
| 169 | |
| 170 | /// @} |
| 171 | |
| 172 | private: |
| 173 | #if GTSAM_ENABLE_BOOST_SERIALIZATION |
| 174 | /** Serialization function */ |
| 175 | friend class boost::serialization::access; |
| 176 | template<class ARCHIVE> |
| 177 | void serialize(ARCHIVE & ar, const unsigned int /*version*/) { |
| 178 | ar & BOOST_SERIALIZATION_BASE_OBJECT_NVP(Base); |
| 179 | } |
| 180 | #endif |
| 181 | }; |
| 182 | |
| 183 | // traits |
| 184 | template<> struct traits<DiscreteBayesNet> : public Testable<DiscreteBayesNet> {}; |
| 185 | |
| 186 | } // \ namespace gtsam |
| 187 | |
| 188 | |