Main Page | Namespace List | Class Hierarchy | Class List | File List | Namespace Members | Class Members | File Members | Related Pages

lemur::distrib::MultiRegrMergeMethod Class Reference

Implementation of regression merging method for single database type case. More...

#include <MultiRegrMergeMethod.hpp>

Inheritance diagram for lemur::distrib::MultiRegrMergeMethod:

lemur::distrib::DistMergeMethod List of all members.

Public Member Functions

 MultiRegrMergeMethod (double constA=0.714, double constB=0.2857)
 the default values of constA and constB are the values of CORI merging algorithm

 ~MultiRegrMergeMethod ()
void mergeScoreSet (const lemur::api::IndexedRealVector &indexset, const DocScoreVector *const *scoreset, DocScoreVector &results)
void calcRegrParams (const lemur::api::IndexedRealVector &indexset, const DocScoreVector *centralscores, const DocScoreVector *const *scoresset)

Protected Member Functions

virtual double score (double dbscore, double docscore, int dbIndex) const
 create a score for the merge based on the index ranking score and the document score

virtual double score (double dbscore, double docscore) const
 create a score for the merge based on the index ranking score and the document score


Protected Attributes

double * parama
 The model parameters.

double * paramb
double defParamA
double defParamB

Detailed Description

Implementation of regression merging method for single database type case.

(All the individual databases are using Inquery(CORI) retrieval algorithm. Build a single regression model for all the databases. When there is not enough training data, the algorithm backs off to CORI results merging algorithm. Assume all the documents scores have been normalized See RetrievalMethod for database ranking algorithms (CORIRetMethod). See DistSearchMethod for multiple individual database retrieval. See CORIMergeMethod for CORI results merging algorithm


Constructor & Destructor Documentation

lemur::distrib::MultiRegrMergeMethod::MultiRegrMergeMethod double  constA = 0.714,
double  constB = 0.2857
 

the default values of constA and constB are the values of CORI merging algorithm

lemur::distrib::MultiRegrMergeMethod::~MultiRegrMergeMethod  ) 
 


Member Function Documentation

void lemur::distrib::MultiRegrMergeMethod::calcRegrParams const lemur::api::IndexedRealVector indexset,
const DocScoreVector centralscores,
const DocScoreVector *const *  scoresset
 

void lemur::distrib::MultiRegrMergeMethod::mergeScoreSet const lemur::api::IndexedRealVector indexset,
const DocScoreVector *const *  scoreset,
DocScoreVector results
[virtual]
 

calculate the single regression model for all the databases indexset are the database scores for selected databases centralsocres are the central documents scores retrieved by centralized sampling database scoresset are the distributed documents scores retrieved by individual databases

Reimplemented from lemur::distrib::DistMergeMethod.

double lemur::distrib::MultiRegrMergeMethod::score double  dbscore,
double  docscore
const [protected, virtual]
 

create a score for the merge based on the index ranking score and the document score

Implements lemur::distrib::DistMergeMethod.

double lemur::distrib::MultiRegrMergeMethod::score double  dbscore,
double  docscore,
int  dbIndex
const [protected, virtual]
 

create a score for the merge based on the index ranking score and the document score


Member Data Documentation

double lemur::distrib::MultiRegrMergeMethod::defParamA [protected]
 

double lemur::distrib::MultiRegrMergeMethod::defParamB [protected]
 

double* lemur::distrib::MultiRegrMergeMethod::parama [protected]
 

The model parameters.

double* lemur::distrib::MultiRegrMergeMethod::paramb [protected]
 


The documentation for this class was generated from the following files:
Generated on Tue Jun 15 11:03:05 2010 for Lemur by doxygen 1.3.4