opennn::BackPropagationLM Struct Reference

Backprop scratch state specific to Levenberg-Marquardt (per-sample errors, Jacobian, Hessian approx). More...

#include <levenberg_marquardt_algorithm.h>

Public Member Functions
	BackPropagationLM (const Index=0, Loss *=nullptr)
	Constructs LM backprop state sized for the given sample count and loss.
virtual	~BackPropagationLM ()=default
void	set (const Index=0, Loss *=nullptr)
	Resizes the LM backprop state for the given sample count and loss.

Public Attributes
Index	samples_number = 0
VectorR	output_deltas
Shape	output_delta_dimensions
Loss *	loss = nullptr
float	error
float	regularization = 0.0f
float	loss_value = 0.0f
VectorR	errors
VectorR	squared_errors
MatrixR	squared_errors_jacobian
VectorR	gradient
MatrixR	hessian

Detailed Description

Backprop scratch state specific to Levenberg-Marquardt (per-sample errors, Jacobian, Hessian approx).

Constructor & Destructor Documentation

BackPropagationLM()

opennn::BackPropagationLM::BackPropagationLM	(	const Index	= 0,
		Loss *	= nullptr )

Constructs LM backprop state sized for the given sample count and loss.

~BackPropagationLM()

virtual opennn::BackPropagationLM::~BackPropagationLM ( )

virtualdefault

Member Function Documentation

set()

void opennn::BackPropagationLM::set	(	const Index	= 0,
		Loss *	= nullptr )

Resizes the LM backprop state for the given sample count and loss.

Member Data Documentation

error

float opennn::BackPropagationLM::error

errors

VectorR opennn::BackPropagationLM::errors

gradient

VectorR opennn::BackPropagationLM::gradient

hessian

MatrixR opennn::BackPropagationLM::hessian

loss

Loss* opennn::BackPropagationLM::loss = nullptr

loss_value

float opennn::BackPropagationLM::loss_value = 0.0f

output_delta_dimensions

Shape opennn::BackPropagationLM::output_delta_dimensions

output_deltas

VectorR opennn::BackPropagationLM::output_deltas

regularization

float opennn::BackPropagationLM::regularization = 0.0f

samples_number

Index opennn::BackPropagationLM::samples_number = 0

squared_errors

VectorR opennn::BackPropagationLM::squared_errors

squared_errors_jacobian

MatrixR opennn::BackPropagationLM::squared_errors_jacobian