Optimizer encapsulates a collection of optimization algorithms used in the toolbox. More...

#include "include/Optimizer.h"

Public Member Functions
	Optimizer (IProcess *process)
	The constructor. More...

void	SGD (const double &gamma0, const unsigned &ini_max_iter, const std::vector< Sequence > &data)
	Stochastic gradient descend. More...

void	PLBFGS (const double &LB, const double &UB)
	Projected LBFGS. More...

void	ProximalGroupLasso (const double &gamma0, const double &lambda, const unsigned &ini_max_iter, const unsigned &group_size)
	Proximal method with group lasso type of regularization. More...

void	ProximalGroupLassoForHawkes (const double &gamma0, const double &lambda, const unsigned &ini_max_iter, const unsigned &group_size)
	Proximal method with group lasso type of regularization for fitting Hawkes processes. More...

void	ProximalNuclear (const double &lambda, const double &rho, const unsigned &ini_max_iter, const Eigen::VectorXd &trueparameters)
	Proximal method with nuclear norm regularization. More...

void	ProximalFrankWolfe (const double &gamma0, const double &lambda, const double &rho, const double &ub_alpha, const unsigned &ini_max_iter, const Eigen::VectorXd &trueparameters)
	Proximal method using conditional gradient update. More...

void	ProximalFrankWolfe (const double &gamma0, const double &lambda, const double &rho, const double &ub_alpha, const unsigned &ini_max_iter)
	Proximal method using conditional gradient update. More...

void	ProximalFrankWolfeForLowRankHawkes (const double &gamma0, const double &lambda0, const double &lambda, const double &ub_lambda0, const double &ub_alpha, const double &rho, const unsigned &ini_max_iter, const unsigned &num_rows, const unsigned &num_cols, const Eigen::MatrixXd &TrueLambda0, const Eigen::MatrixXd &TrueAlpha)
	Proximal method using conditional gradient update for Low-rank Hawkes process. More...

void	ProximalFrankWolfeForLowRankHawkes (const double &gamma0, const double &lambda0, const double &lambda, const double &ub_lambda0, const double &ub_alpha, const double &rho, const unsigned &ini_max_iter, const unsigned &num_rows, const unsigned &num_cols)
	Proximal method using conditional gradient update for Low-rank Hawkes process. More...

Detailed Description

Optimizer encapsulates a collection of optimization algorithms used in the toolbox.

Definition at line 17 of file Optimizer.h.

Constructor & Destructor Documentation

Optimizer::Optimizer ( IProcess * process )

inline

The constructor.

Parameters

process the given point process model to fit.

Definition at line 90 of file Optimizer.h.

Member Function Documentation

void Optimizer::PLBFGS	(	const double &	LB,
		const double &	UB
	)

Projected LBFGS.

Parameters

[in]	LB	lower bound of the parameters.
[in]	UB	upper bound of the parameters.

Definition at line 230 of file Optimizer.cc.

void Optimizer::ProximalFrankWolfe	(	const double &	gamma0,
		const double &	lambda,
		const double &	rho,
		const double &	ub_alpha,
		const unsigned &	ini_max_iter,
		const Eigen::VectorXd &	trueparameters
	)

Proximal method using conditional gradient update.

Parameters

gamma0	initial learning rate.
lambda	regularization coefficient.
rho	coefficient to enforce the low-rank constraint.
ub_alpha	upper bound of the nuclear norm of the excitation matrix.
ini_max_iter	initial maximum number of iterations.
trueparameters	true model parameters to compare with the current solution.

Definition at line 709 of file Optimizer.cc.

void Optimizer::ProximalFrankWolfe	(	const double &	gamma0,
		const double &	lambda,
		const double &	rho,
		const double &	ub_alpha,
		const unsigned &	ini_max_iter
	)

Proximal method using conditional gradient update.

Parameters

gamma0	initial learning rate.
lambda	regularization coefficient.
rho	coefficient to enforce the low-rank constraint.
ub_alpha	upper bound of the nuclear norm of the excitation matrix.
ini_max_iter	initial maximum number of iterations.

Definition at line 785 of file Optimizer.cc.

void Optimizer::ProximalFrankWolfeForLowRankHawkes	(	const double &	gamma0,
		const double &	lambda0,
		const double &	lambda,
		const double &	ub_lambda0,
		const double &	ub_alpha,
		const double &	rho,
		const unsigned &	ini_max_iter,
		const unsigned &	num_rows,
		const unsigned &	num_cols,
		const Eigen::MatrixXd &	TrueLambda0,
		const Eigen::MatrixXd &	TrueAlpha
	)

Proximal method using conditional gradient update for Low-rank Hawkes process.

Parameters

gamma0	initial learning rate.
lambda0	regularization coefficient for the base intensity function.
lambda	regularization coefficient for the excitation matrix.
ub_lambda0	upper bound of the nuclear norm of the base intensity matrix.
ub_alpha	upper bound of the nuclear norm of the excitation matrix.
rho	coefficient to enforce the low-rank constraint.
ini_max_iter	initial maximum number of iterations.
num_rows	the number of rows (or users) of the base intensity (or excitation) matrix.
num_cols	the number of cols (or items) of the base intensity (or excitation) matrix.
TrueLambda0	true model parameters of the base intensity matrix to compare with the current solution.
TrueAlpha	true model parameters of the excitation matrix to compare with the current solution.

Definition at line 861 of file Optimizer.cc.

void Optimizer::ProximalFrankWolfeForLowRankHawkes	(	const double &	gamma0,
		const double &	lambda0,
		const double &	lambda,
		const double &	ub_lambda0,
		const double &	ub_alpha,
		const double &	rho,
		const unsigned &	ini_max_iter,
		const unsigned &	num_rows,
		const unsigned &	num_cols
	)

Proximal method using conditional gradient update for Low-rank Hawkes process.

Parameters

gamma0	initial learning rate.
lambda0	regularization coefficient for the base intensity function.
lambda	regularization coefficient for the excitation matrix.
ub_lambda0	upper bound of the nuclear norm of the base intensity matrix.
ub_alpha	upper bound of the nuclear norm of the excitation matrix.
rho	coefficient to enforce the low-rank constraint.
ini_max_iter	initial maximum number of iterations.
num_rows	the number of rows (or users) of the base intensity (or excitation) matrix.
num_cols	the number of cols (or items) of the base intensity (or excitation) matrix.

Definition at line 953 of file Optimizer.cc.

void Optimizer::ProximalGroupLasso	(	const double &	gamma0,
		const double &	lambda,
		const unsigned &	ini_max_iter,
		const unsigned &	group_size
	)

Proximal method with group lasso type of regularization.

Parameters

gamma0	initial learning rate.
lambda	regularization coefficient.
ini_max_iter	initial maximum number of iterations.
group_size	the size of the entire parameter vector to be pushed into zero.

Definition at line 454 of file Optimizer.cc.

void Optimizer::ProximalGroupLassoForHawkes	(	const double &	gamma0,
		const double &	lambda,
		const unsigned &	ini_max_iter,
		const unsigned &	group_size
	)

Proximal method with group lasso type of regularization for fitting Hawkes processes.

Parameters

gamma0	initial learning rate.
lambda	regularization coefficient.
ini_max_iter	initial maximum number of iterations.
group_size	the size of the entire parameter vector to be pushed into zero.

Definition at line 579 of file Optimizer.cc.

void Optimizer::ProximalNuclear	(	const double &	lambda,
		const double &	rho,
		const unsigned &	ini_max_iter,
		const Eigen::VectorXd &	trueparameters
	)

Proximal method with nuclear norm regularization.

Parameters

lambda	regularization coefficient.
rho	coefficient to enforce the low-rank constraint.
ini_max_iter	initial maximum number of iterations.
trueparameters	true model parameters to compare with the current solution.

Definition at line 639 of file Optimizer.cc.

void Optimizer::SGD	(	const double &	gamma0,
		const unsigned &	ini_max_iter,
		const std::vector< Sequence > &	data
	)

Stochastic gradient descend.

Parameters

[in]	gamma0	initial learning rate.
[in]	ini_max_iter	initial maximum number of iterations.
[in]	data	the sampled sequence of events.

Definition at line 14 of file Optimizer.cc.

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

/Users/nandu/Development/MultiVariatePointProcess/include/Optimizer.h
/Users/nandu/Development/MultiVariatePointProcess/src/Optimizer.cc

Public Member Functions

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation