ldaplusplus::em::MStepInterface< Scalar > Class Template Reference

#include <MStepInterface.hpp>

Inheritance diagram for ldaplusplus::em::MStepInterface< Scalar >:

Public Member Functions
virtual void	m_step (std::shared_ptr< parameters::Parameters > parameters)=0
virtual void	doc_m_step (const std::shared_ptr< corpus::Document > doc, const std::shared_ptr< parameters::Parameters > v_parameters, std::shared_ptr< parameters::Parameters > m_parameters)=0
Public Member Functions inherited from ldaplusplus::events::EventDispatcherComposition
std::shared_ptr< EventDispatcherInterface >	get_event_dispatcher ()
void	set_event_dispatcher (std::shared_ptr< EventDispatcherInterface > dispatcher)

Detailed Description

template<typename Scalar>
class ldaplusplus::em::MStepInterface< Scalar >

Interface that defines an M-step iteration for any LDA inference.

The maximization step maximizes the likelihood (actually the Evidence Lower Bound) of the data by changing the parameters and using the variational parameters as constants. In classical LDA this step computes the distribution over words for all topics using the variational parameters \(\phi\) and \(\gamma\).

Member Function Documentation

template<typename Scalar >

virtual void ldaplusplus::em::MStepInterface< Scalar >::doc_m_step ( const std::shared_ptr< corpus::Document >  doc, const std::shared_ptr< parameters::Parameters >  v_parameters, std::shared_ptr< parameters::Parameters >  m_parameters  )

pure virtual

Perform calculations for a specific document.

The variational parameters are only passed to the maximization step in this method. In other implementations this method is usually called sufficient statistics.

This method allows for the implementation of online LDA inference methods.

Parameters

doc	A single document
v_parameters	The variational parameters computed in the e-step
m_parameters	Model parameters could be changed in case of online methods

Implemented in ldaplusplus::em::FastOnlineSupervisedMStep< Scalar >, ldaplusplus::em::FastSupervisedMStep< Scalar >, ldaplusplus::em::SupervisedMStep< Scalar >, ldaplusplus::em::UnsupervisedMStep< Scalar >, ldaplusplus::em::CorrespondenceSupervisedMStep< Scalar >, ldaplusplus::em::SemiSupervisedMStep< Scalar >, and ldaplusplus::em::MultinomialSupervisedMStep< Scalar >.

template<typename Scalar >

virtual void ldaplusplus::em::MStepInterface< Scalar >::m_step ( std::shared_ptr< parameters::Parameters >  parameters )

pure virtual

Maximize the ELBO.

This function usually changes the passed in parameters.

Parameters

parameters

Model parameters (maybe changed after call)

Implemented in ldaplusplus::em::FastOnlineSupervisedMStep< Scalar >, ldaplusplus::em::FastSupervisedMStep< Scalar >, ldaplusplus::em::SupervisedMStep< Scalar >, ldaplusplus::em::UnsupervisedMStep< Scalar >, ldaplusplus::em::CorrespondenceSupervisedMStep< Scalar >, and ldaplusplus::em::MultinomialSupervisedMStep< Scalar >.

---

The documentation for this class was generated from the following file:

• include/ldaplusplus/em/MStepInterface.hpp