* Introduced CCNODEBUGFLAGS in Makefile
* Updated copyright headers
* Added max-product functionality to JTree
-* [Charlie Vaske] EM bugfix (was using abs() instead of fabs() in
+* [Charles Vaske] EM bugfix (was using abs() instead of fabs() in
determining EM termination, which caused a loss of precision).
-* [Charlie Vaske] New constructor in Permute for canonical variable ordering
-* [Charlie Vaske] New constructor in Factor that reorders variables to canonical ordering
-* [Charlie Vaske] New accessor in Properties for getting all keys
-* [Charlie Vaske] New function for getting inferred parameters
+* [Charles Vaske] New constructor in Permute for canonical variable ordering
+* [Charles Vaske] New constructor in Factor that reorders variables to canonical ordering
+* [Charles Vaske] New accessor in Properties for getting all keys
+* [Charles Vaske] New function for getting inferred parameters
* Cleaned up error handling by introducing the DAI_THROWE macro
* Introduced DAI_DEBASSERT macro to abbreviate a common pattern
* Changed FactorGraph::clamp and DAIAlg::clamp interfaces (the variable to be
* Added BipartiteGraph::eraseEdge
* Removed all the virtual default constructors *::create(), as they are not used.
* Fixed bug in MaxSpanningTree (it wrongly assumed that the tree was not empty).
-* [Charlie Vaske] Added Expectation Maximization code.
+* [Charles Vaske] Added Expectation Maximization code.
* Added MatLab QuickStart to README.
* MEX file dai now also returns variable and factor beliefs.
* Cleanups and extra documentation of the code contributed by Frederik Eaton
/// \file
-/// \brief Defines class BP_dual
-/// \todo This replicates a large part of the functionality of BP; would it not be shorter to adapt BP instead?
-/// \todo Improve documentation
+/// \brief Defines class BP_dual, which is used primarily by BBP.
+/// \todo This replicates a large part of the functionality of BP; would it not be more efficient to adapt BP instead?
+/// \author Frederik Eaton
#ifndef __defined_libdai_bp_dual_h
* to variables and messages from variables to factors), and normalizers, given an InfAlg.
* These are computed from the variable and factor beliefs of the InfAlg.
* This class is used primarily by BBP.
+ *
+ * \author Frederik Eaton
*/
class BP_dual {
-
protected:
/// Convenience label for storing edge properties
template<class T>
/// Does all necessary preprocessing
void init();
- /// Allocates space for _msgs
+ /// Allocates space for \a _msgs
void regenerateMessages();
- /// Allocates space for _beliefs
+ /// Allocates space for \a _beliefs
void regenerateBeliefs();
/// Calculate all messages from InfAlg beliefs
void calcMessages();
- /// Update factor->variable message (i->I)
+ /// Update factor->variable message (\a i -> \a I)
void calcNewM(size_t i, size_t _I);
- /// Update variable->factor message (I->i)
+ /// Update variable->factor message (\a I -> \a i)
void calcNewN(size_t i, size_t _I);
/// Calculate all variable and factor beliefs from messages
void calcBeliefs();
- /// Calculate variable belief
+ /// Calculate belief of variable \a i
void calcBeliefV(size_t i);
- /// Calculate factor belief
+ /// Calculate belief of factor \a I
void calcBeliefF(size_t I);
public:
/// Construct BP_dual object from (converged) InfAlg object's beliefs and factors.
- /* A pointer to the the InfAlg object is stored,
+ /** \warning A pointer to the the InfAlg object is stored,
* so the object must not be destroyed before the BP_dual is destroyed.
*/
BP_dual( const InfAlg *ia ) : _ia(ia) { init(); }
/// Returns the underlying FactorGraph
const FactorGraph& fg() const { return _ia->fg(); }
- /// Returns reference to factor -> var message (I->i)
+ /// Returns reference to factor->variable message (\a I -> \a i)
Prob & msgM( size_t i, size_t _I ) { return _msgs.m[i][_I]; }
- /// Returns constant reference to factor -> var message (I->i)
+ /// Returns constant reference to factor->variable message (\a I -> \a i)
const Prob & msgM( size_t i, size_t _I ) const { return _msgs.m[i][_I]; }
- /// Returns reference to var -> factor message (i->I)
+ /// Returns reference to variable -> factor message (\a i -> \a I)
Prob & msgN( size_t i, size_t _I ) { return _msgs.n[i][_I]; }
- /// Returns constant reference to var -> factor message (i->I)
+ /// Returns constant reference to variable -> factor message (\a i -> \a I)
const Prob & msgN( size_t i, size_t _I ) const { return _msgs.n[i][_I]; }
- /// Returns reference to normalizer for msgM
+ /// Returns reference to normalizer for factor->variable message (\a I -> \a i)
Real & zM( size_t i, size_t _I ) { return _msgs.Zm[i][_I]; }
- /// Returns constant reference to normalizer for msgM
+ /// Returns constant reference to normalizer for factor->variable message (\a I -> \a i)
const Real & zM( size_t i, size_t _I ) const { return _msgs.Zm[i][_I]; }
- /// Returns reference to normalizer for msgN
+ /// Returns reference to normalizer for variable -> factor message (\a i -> \a I)
Real & zN( size_t i, size_t _I ) { return _msgs.Zn[i][_I]; }
- /// Returns constant reference to normalizer for msgN
+ /// Returns constant reference to normalizer for variable -> factor message (\a i -> \a I)
const Real & zN( size_t i, size_t _I ) const { return _msgs.Zn[i][_I]; }
- /// Returns variable belief
+ /// Returns belief of variable \a i
Factor beliefV( size_t i ) const { return Factor( _ia->fg().var(i), _beliefs.b1[i] ); }
- /// Returns factor belief
+ /// Returns belief of factor \a I
Factor beliefF( size_t I ) const { return Factor( _ia->fg().factor(I).vars(), _beliefs.b2[I] ); }
- /// Returns normalizer for variable belief
+ /// Returns normalizer for belief of variable \a i
Real beliefVZ( size_t i ) const { return _beliefs.Zb1[i]; }
- /// Returns normalizer for factor belief
+ /// Returns normalizer for belief of factor \a I
Real beliefFZ( size_t I ) const { return _beliefs.Zb2[I]; }
};
/// \brief Defines classes related to Expectation Maximization: EMAlg, ParameterEstimation, CondProbEstimation and SharedParameters
/// \todo Describe EM file format
/// \todo Improve documentation
+/// \author Charles Vaske
namespace dai {
* via the static ParameterEstimation::registerMethod function.
* The default registry only contains CondProbEstimation, named
* "ConditionalProbEstimation".
+ *
+ * \author Charles Vaske
*/
class ParameterEstimation {
public:
/// Estimates the parameters of a conditional probability table, using pseudocounts.
+/** \author Charles Vaske
+ */
class CondProbEstimation : private ParameterEstimation {
private:
/// Number of states of the variable of interest
* class couples one or more factors (together with the specified orderings
* of the variables) with a ParameterEstimation object, taking care of the
* necessary permutations of the factor entries / parameters.
+ *
+ * \author Charles Vaske
*/
class SharedParameters {
public:
/// A MaximizationStep groups together several parameter estimation tasks into a single unit.
+/** \author Charles Vaske
+ */
class MaximizationStep {
private:
std::vector<SharedParameters> _params;
* Having multiple and separate maximization steps allows for maximizing some
* parameters, performing another E step, and then maximizing separate
* parameters, which may result in faster convergence in some cases.
+ *
+ * \author Charles Vaske
*/
class EMAlg {
private:
projects / libdai.git / commitdiff