WeightUpdater Class Reference

Weight and Lambda update heuristic. More...

#include <WeightUpdater.h>

Public Member Functions
	WeightUpdater (AlternativesPerformance &ap, Config &conf)
	WeightUpdater (const WeightUpdater &wu)
void	updateWeightsAndLambda (MRSortModel &model)
std::vector< std::vector< std::vector< bool > > >	computeXMatrix (MRSortModel &model)
std::vector< std::vector< std::vector< bool > > >	computeYMatrix (MRSortModel &model)
bool	modelCheck (MRSortModel &model)

Detailed Description

Weight and Lambda update heuristic.

This class implements the update of Criterion weight and the majority threshold lambda as described in the thesis. This algorithm is initialized with the dataset and can then be used to transform a model into an updated one.

The solving of the linear problem is done externally by the LinearSolver. The WeightUpdater class is responsible for computing the constraint matrix given a specific model and passing them to the linear solver.

A complete description of the heuristic can be found in Lambda and Weight updater.

Constructor & Destructor Documentation

WeightUpdater() [1/2]

WeightUpdater::WeightUpdater	(	AlternativesPerformance &	ap,
		Config &	conf
	)

WeightUpdater standard constructor.

Parameters

ap	AlternativesPerformance objet that represents the dataset of the problem
config	config setup of the app

WeightUpdater() [2/2]

WeightUpdater::WeightUpdater

(

const WeightUpdater &

wu

)

WeightUpdater constructor by copy.

Parameters

wu	WeightUpdater objet to copy.

Member Function Documentation

computeXMatrix()

std::vector<std::vector<std::vector<bool> > > WeightUpdater::computeXMatrix

(

MRSortModel &

model

)

computexMatrixX Computes linear constraint matrix for x variables. X Constraint matrix are of dimension (n_prof - 1, n_alt, n_crit). In the first dimension, we have the n_prof - 1 profiles of the problem. Inside profile h, we have a row for all alternatives, empty if not classified in profile h. The representation of an alternative a classified in profile h represents the presence or absence of each criterion in the constraint (satisfied as a_i >= b_h-1_i for i criterion).

Parameters

model

MRSortModel use to compute the x matrix

Returns

x_matrix matrix recapitulating the constraints to add to the linear problem for the x variables

computeYMatrix()

std::vector<std::vector<std::vector<bool> > > WeightUpdater::computeYMatrix

(

MRSortModel &

model

)

computexMatrixY Computes linear constraint matrix for y variables. Y Constraint matrix are of dimension (n_prof - 1, n_alt, n_crit). In the first dimension, we have the n_prof - 1 profiles of the problem. Inside profile h, we have a row for all alternatives, empty if not classified in profile h. The representation of an alternative a classified in profile h represents the presence or absence of each criterion in the constraint (satisfied as a_i >= b_h_i for i criterion).

Parameters

model

MRSortModel use to compute the y matrix

Returns

y_matrix matrix recapitulating the constraints to add to the linear problem for the y variables

modelCheck()

bool WeightUpdater::modelCheck

(

MRSortModel &

model

)

modelCheck Checks if the profile is suited to be updated with this weight updater: it checks if the criterion in the profile appears in the same order as in the dataset.

Parameters

model

MRSortModel to check

Returns

is_suited

updateWeightsAndLambda()

void WeightUpdater::updateWeightsAndLambda

(

MRSortModel &

model

)

updateWeightsAndLambda implements the metaheuristic that updates the weights and the majority threshold given the state of the profiles in the model.

Parameters

model

MRSortModel to update

---

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

• /Users/ThomasVindard 1/Desktop/3A/Projet/fastPL/include/learning/WeightUpdater.h