Weighted smooth-1norm activation. More...

#include <weighted-smooth-1norm.hpp>

Inheritance diagram for ActivationModelWeightedSmooth1NormTpl< _Scalar >:

Public Types
typedef ActivationDataAbstractTpl< Scalar >	ActivationDataAbstract
typedef ActivationModelAbstractTpl< Scalar >	Base
typedef ActivationDataWeightedSmooth1NormTpl< Scalar >	Data
typedef MathBaseTpl< Scalar >	MathBase
typedef MathBase::VectorXs	VectorXs
Public Types inherited from ActivationModelAbstractTpl< _Scalar >
typedef ActivationDataAbstractTpl< Scalar >	ActivationDataAbstract
typedef MathBaseTpl< Scalar >	MathBase
typedef MathBase::MatrixXs	MatrixXs
typedef MathBase::VectorXs	VectorXs

Public Member Functions
	ActivationModelWeightedSmooth1NormTpl (const VectorXs &weights, const Scalar delta=Scalar(1.))
	Initialize the weighted smooth-1norm activation model.
virtual void	calc (const std::shared_ptr< ActivationDataAbstract > &data, const Eigen::Ref< const VectorXs > &r) override
	Compute the weighted smooth-1norm activation.
virtual void	calcDiff (const std::shared_ptr< ActivationDataAbstract > &data, const Eigen::Ref< const VectorXs > &r) override
	Compute the activation derivatives.
template<typename NewScalar>
ActivationModelWeightedSmooth1NormTpl< NewScalar >	cast () const
virtual std::shared_ptr< ActivationDataAbstract >	createData () override
	Create the weighted smooth-1norm activation data.
EIGEN_MAKE_ALIGNED_OPERATOR_NEW	CROCODDYL_DERIVED_CAST (ActivationModelBase, ActivationModelWeightedSmooth1NormTpl) typedef _Scalar Scalar
Scalar	get_delta () const
const VectorXs &	get_weights () const
virtual void	print (std::ostream &os) const override
	Print relevant information of the weighted smooth-1norm model.
void	set_weights (const VectorXs &weights)
Public Member Functions inherited from ActivationModelAbstractTpl< _Scalar >
	ActivationModelAbstractTpl (const std::size_t nr)
std::size_t	get_nr () const

Protected Attributes
std::size_t	nr_
Protected Attributes inherited from ActivationModelAbstractTpl< _Scalar >
std::size_t	nr_

Additional Inherited Members
Public Attributes inherited from ActivationModelAbstractTpl< _Scalar >
EIGEN_MAKE_ALIGNED_OPERATOR_NEW typedef _Scalar	Scalar

Detailed Description

template<typename _Scalar>
class crocoddyl::ActivationModelWeightedSmooth1NormTpl< _Scalar >

Weighted smooth-1norm activation.

This activation function describes a weighted smooth representation of the 1-norm of a residual vector:

$\‍[ a(\mathbf{r}) = \sum_{i=0}^{nr-1} w_i\delta^2\left(\sqrt{1 + (r_i/\delta)^2} - 1\right), \‍]$

where $\delta > 0$ is the smoothing scale, $w_i \geq 0$ is the weight associated with the scalar residual , and is the residual dimension. This is the weighted classical pseudo-Huber form, equivalently $w_i(\delta\sqrt{\delta^2 + r_i^2} - \delta^2)$ .

The weights scale the activation without changing the smoothing transition, which occurs around $|r_i| = \delta$ . The local quadratic approximation is for every $\delta$ , while the asymptotic slope of each term is $w_i\delta$ .

See also: calc(), calcDiff(), createData()

Definition at line 40 of file weighted-smooth-1norm.hpp.

Member Typedef Documentation

◆ MathBase

template<typename _Scalar>

typedef MathBaseTpl<Scalar> MathBase

Definition at line 48 of file weighted-smooth-1norm.hpp.

◆ Base

template<typename _Scalar>

typedef ActivationModelAbstractTpl<Scalar> Base

Definition at line 49 of file weighted-smooth-1norm.hpp.

◆ ActivationDataAbstract

template<typename _Scalar>

typedef ActivationDataAbstractTpl<Scalar> ActivationDataAbstract

Definition at line 50 of file weighted-smooth-1norm.hpp.

◆ Data

template<typename _Scalar>

typedef ActivationDataWeightedSmooth1NormTpl<Scalar> Data

Definition at line 51 of file weighted-smooth-1norm.hpp.

◆ VectorXs

template<typename _Scalar>

typedef MathBase::VectorXs VectorXs

Definition at line 52 of file weighted-smooth-1norm.hpp.

Constructor & Destructor Documentation

◆ ActivationModelWeightedSmooth1NormTpl()

template<typename _Scalar>

ActivationModelWeightedSmooth1NormTpl	(	const VectorXs &	weights,
		const Scalar	delta = Scalar(1.) )

inlineexplicit

Initialize the weighted smooth-1norm activation model.

Parameters

[in]	weights	Nonnegative residual weights
[in]	delta	Strictly positive smoothing scale (default: 1)

Definition at line 60 of file weighted-smooth-1norm.hpp.

Member Function Documentation

◆ calc()

template<typename _Scalar>

virtual void calc	(	const std::shared_ptr< ActivationDataAbstract > &	data,
		const Eigen::Ref< const VectorXs > &	r )

inlineoverridevirtual

Compute the weighted smooth-1norm activation.

Parameters

[in]	data	Weighted smooth-1norm activation data
[in]	r	Residual vector $\mathbf{r}\in\mathbb{R}^{nr}$

Implements ActivationModelAbstractTpl< _Scalar >.

Definition at line 77 of file weighted-smooth-1norm.hpp.

◆ calcDiff()

template<typename _Scalar>

virtual void calcDiff	(	const std::shared_ptr< ActivationDataAbstract > &	data,
		const Eigen::Ref< const VectorXs > &	r )

inlineoverridevirtual

Compute the activation derivatives.

Parameters

[in]	data	Weighted smooth-1norm activation data
[in]	r	Residual vector $\mathbf{r}\in\mathbb{R}^{nr}$

Implements ActivationModelAbstractTpl< _Scalar >.

Definition at line 98 of file weighted-smooth-1norm.hpp.

◆ createData()

template<typename _Scalar>

virtual std::shared_ptr< ActivationDataAbstract > createData ( )

inlineoverridevirtual

Create the weighted smooth-1norm activation data.

Returns: the activation data

Reimplemented from ActivationModelAbstractTpl< _Scalar >.

Definition at line 118 of file weighted-smooth-1norm.hpp.

◆ cast()

template<typename _Scalar>

template<typename NewScalar>

ActivationModelWeightedSmooth1NormTpl< NewScalar > cast ( ) const

inline

Definition at line 123 of file weighted-smooth-1norm.hpp.

◆ get_weights()

template<typename _Scalar>

const VectorXs & get_weights ( ) const

inline

Definition at line 129 of file weighted-smooth-1norm.hpp.

◆ get_delta()

template<typename _Scalar>

Scalar get_delta ( ) const

inline

Definition at line 130 of file weighted-smooth-1norm.hpp.

◆ set_weights()

template<typename _Scalar>

void set_weights ( const VectorXs & weights )

inline

Definition at line 132 of file weighted-smooth-1norm.hpp.

◆ print()

template<typename _Scalar>

virtual void print ( std::ostream & os ) const

inlineoverridevirtual

Print relevant information of the weighted smooth-1norm model.

Parameters

[out] os Output stream object

Reimplemented from ActivationModelAbstractTpl< _Scalar >.

Definition at line 147 of file weighted-smooth-1norm.hpp.

Member Data Documentation

◆ nr_

template<typename _Scalar>

std::size_t nr_

protected

Definition at line 68 of file activation-base.hpp.

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

/build/source/include/crocoddyl/core/activations/weighted-smooth-1norm.hpp

Public Types

Public Member Functions

Protected Attributes

Additional Inherited Members

Detailed Description

Member Typedef Documentation

◆ MathBase

◆ Base

◆ ActivationDataAbstract

◆ Data

◆ VectorXs

Constructor & Destructor Documentation

◆ ActivationModelWeightedSmooth1NormTpl()

Member Function Documentation

◆ calc()

◆ calcDiff()

◆ createData()

◆ cast()

◆ get_weights()

◆ get_delta()

◆ set_weights()

◆ print()

Member Data Documentation

◆ nr_