netdem::RegressionNet Class Reference

A class that represents a feedforward neural network for regression. More...

#include <regression_net.hpp>

Public Member Functions
void	ResetModel ()
	Resets the neural network model.
void	AddLayer (MLPackUtils::LayerType layer_name,...)
	Adds a layer to the neural network model.
void	Train (const arma::mat &data_x, const arma::mat &data_y)
	Trains the neural network model with data.
arma::mat	Predict (const arma::mat &data_x)
	Predicts with the neural network model using input data.
arma::mat	Gradient (const arma::mat &data_x)
	Calculates the gradient of the neural network model with respect to input data.
void	Load (std::string const &filename, std::string const &label)
	Loads the neural network model from disk.
void	Save (std::string const &filename, std::string const &label)
	Saves the neural network model to disk.

Public Attributes
mlpack::FFN< mlpack::MeanSquaredError, mlpack::HeInitialization >	model
	The model object that represents the neural network.
double	step_size {0.01}
	The step size used for optimization during training.
int	batch_size {32}
	The batch size used for mini-batch gradient descent during training.
double	decay_rate_moment {0.9}
	The decay rate used for momentum during training.
double	decay_rate_norm {0.9}
	The decay rate used for weight normalization during training.
double	gradient_init_param {1e-8}
	Initial parameter value used for weight initialization during training.
int	epochs {100}
	The number of epochs used for training.
double	stop_tol {1e-8}
	The stopping tolerance used for determining when to stop training.
bool	enable_logging {true}
	Whether or not to enable logging during training.

Detailed Description

A class that represents a feedforward neural network for regression.

The RegressionNet class is a specific implementation of a feedforward neural network for regression. It provides functionality to reset the model, add layers to the model, train the model with data, predict with the model, calculate the gradient of the model with respect to the input data, and load/save the model to disk.

Examples

01_train_netsdf.cpp, 02_test_netsdf.cpp, 03_train_netspf.cpp, 04_test_netspf.cpp, 04_train_regressor_trimesh_plane.cpp, 05_test_regressor_trimesh_plane.cpp, 06_test_ann_vs_geom_trimesh_plane.cpp, 11_train_coded_netsdf.cpp, 12_test_coded_netsdf.cpp, 14_train_regressor_trimesh.cpp, 15_test_coded_shape.cpp, 15_test_regressor_trimesh.cpp, 16_test_ann_vs_geom_trimesh.cpp, 16_test_latent_space.cpp, 17_test_ann_vs_geom_trimesh.cpp, 24_train_regressor_ellipsoid_plane.cpp, 25_test_regressor_ellipsoid_plane.cpp, 26_test_ann_vs_geom_ellipsoid_plane.cpp, 34_train_regressor_ellipsoid.cpp, 35_test_regressor_ellipsoid.cpp, 36_test_ann_vs_geom_ellipsoid.cpp, and 37_test_ann_vs_geom_ellipsoid.cpp.

Member Function Documentation

AddLayer()

void netdem::RegressionNet::AddLayer	(	MLPackUtils::LayerType	layer_name,
			... )

Adds a layer to the neural network model.

This function adds a layer to the neural network model.

Parameters

layer_name	The type of layer to add.
...	Additional arguments depending on the type of layer being added.

For MLPackUtils::LayerType::Linear, an additional parameter indicating the dimension of outputs is required.

Examples

01_train_netsdf.cpp, 03_train_netspf.cpp, 04_train_regressor_trimesh_plane.cpp, 11_train_coded_netsdf.cpp, 14_train_regressor_trimesh.cpp, 24_train_regressor_ellipsoid_plane.cpp, and 34_train_regressor_ellipsoid.cpp.

Gradient()

arma::mat netdem::RegressionNet::Gradient

(

const arma::mat &

data_x

)

Calculates the gradient of the neural network model with respect to input data.

This function calculates the gradient of the neural network model with respect to the input data using backpropagation.

Parameters

data_x

The input data matrix.

Returns

The gradient of the neural network model with respect to the input data as a matrix of size (input_dim x num_samples).

Examples

11_train_coded_netsdf.cpp, and 16_test_latent_space.cpp.

Load()

void netdem::RegressionNet::Load	(	std::string const &	filename,
		std::string const &	label )

Loads the neural network model from disk.

This function loads the neural network model from disk.

Parameters

filename	The name of the file containing the model data.
label	The label of the model data within the file (if applicable).

Examples

02_test_netsdf.cpp, 04_test_netspf.cpp, 05_test_regressor_trimesh_plane.cpp, 06_test_ann_vs_geom_trimesh_plane.cpp, 12_test_coded_netsdf.cpp, 15_test_coded_shape.cpp, 15_test_regressor_trimesh.cpp, 16_test_ann_vs_geom_trimesh.cpp, 16_test_latent_space.cpp, 17_test_ann_vs_geom_trimesh.cpp, 25_test_regressor_ellipsoid_plane.cpp, 26_test_ann_vs_geom_ellipsoid_plane.cpp, 35_test_regressor_ellipsoid.cpp, 36_test_ann_vs_geom_ellipsoid.cpp, and 37_test_ann_vs_geom_ellipsoid.cpp.

Predict()

arma::mat netdem::RegressionNet::Predict

(

const arma::mat &

data_x

)

Predicts with the neural network model using input data.

This function predicts with the neural network model using the provided input data.

Parameters

data_x

The input data matrix.

Returns

The predicted output data as a matrix of size (output_dim x num_samples).

Examples

01_train_netsdf.cpp, 02_test_netsdf.cpp, 03_train_netspf.cpp, 04_test_netspf.cpp, 04_train_regressor_trimesh_plane.cpp, 05_test_regressor_trimesh_plane.cpp, 06_test_ann_vs_geom_trimesh_plane.cpp, 11_train_coded_netsdf.cpp, 12_test_coded_netsdf.cpp, 14_train_regressor_trimesh.cpp, 15_test_coded_shape.cpp, 15_test_regressor_trimesh.cpp, 16_test_ann_vs_geom_trimesh.cpp, 16_test_latent_space.cpp, 17_test_ann_vs_geom_trimesh.cpp, 24_train_regressor_ellipsoid_plane.cpp, 25_test_regressor_ellipsoid_plane.cpp, 26_test_ann_vs_geom_ellipsoid_plane.cpp, 34_train_regressor_ellipsoid.cpp, 35_test_regressor_ellipsoid.cpp, 36_test_ann_vs_geom_ellipsoid.cpp, and 37_test_ann_vs_geom_ellipsoid.cpp.

ResetModel()

void netdem::RegressionNet::ResetModel

(

)

Resets the neural network model.

This function resets the neural network model.

Save()

void netdem::RegressionNet::Save	(	std::string const &	filename,
		std::string const &	label )

Saves the neural network model to disk.

This function saves the neural network model to disk.

Parameters

filename	The name of the file to save the model data to.
label	The label of the model data within the file (if applicable).

Examples

01_train_netsdf.cpp, 03_train_netspf.cpp, 04_train_regressor_trimesh_plane.cpp, 11_train_coded_netsdf.cpp, 14_train_regressor_trimesh.cpp, 24_train_regressor_ellipsoid_plane.cpp, and 34_train_regressor_ellipsoid.cpp.

Train()

void netdem::RegressionNet::Train	(	const arma::mat &	data_x,
		const arma::mat &	data_y )

Trains the neural network model with data.

This function trains the neural network model with the provided input and output data using mini-batch gradient descent with momentum and weight normalization.

Parameters

data_x	The input data matrix.
data_y	The output data matrix.

Examples

01_train_netsdf.cpp, 03_train_netspf.cpp, 04_train_regressor_trimesh_plane.cpp, 11_train_coded_netsdf.cpp, 14_train_regressor_trimesh.cpp, 24_train_regressor_ellipsoid_plane.cpp, and 34_train_regressor_ellipsoid.cpp.

Member Data Documentation

batch_size

int netdem::RegressionNet::batch_size {32}

The batch size used for mini-batch gradient descent during training.

Examples

01_train_netsdf.cpp, 03_train_netspf.cpp, 04_train_regressor_trimesh_plane.cpp, 11_train_coded_netsdf.cpp, 14_train_regressor_trimesh.cpp, 24_train_regressor_ellipsoid_plane.cpp, and 34_train_regressor_ellipsoid.cpp.

decay_rate_moment

double netdem::RegressionNet::decay_rate_moment {0.9}

The decay rate used for momentum during training.

Examples

34_train_regressor_ellipsoid.cpp.

decay_rate_norm

double netdem::RegressionNet::decay_rate_norm {0.9}

The decay rate used for weight normalization during training.

Examples

34_train_regressor_ellipsoid.cpp.

enable_logging

bool netdem::RegressionNet::enable_logging {true}

Whether or not to enable logging during training.

Examples

01_train_netsdf.cpp, 03_train_netspf.cpp, and 11_train_coded_netsdf.cpp.

epochs

int netdem::RegressionNet::epochs {100}

The number of epochs used for training.

Examples

01_train_netsdf.cpp, 03_train_netspf.cpp, 04_train_regressor_trimesh_plane.cpp, 11_train_coded_netsdf.cpp, 14_train_regressor_trimesh.cpp, 24_train_regressor_ellipsoid_plane.cpp, and 34_train_regressor_ellipsoid.cpp.

gradient_init_param

double netdem::RegressionNet::gradient_init_param {1e-8}

Initial parameter value used for weight initialization during training.

model

mlpack::FFN<mlpack::MeanSquaredError, mlpack::HeInitialization> netdem::RegressionNet::model

The model object that represents the neural network.

step_size

double netdem::RegressionNet::step_size {0.01}

The step size used for optimization during training.

Examples

01_train_netsdf.cpp, 03_train_netspf.cpp, 04_train_regressor_trimesh_plane.cpp, 11_train_coded_netsdf.cpp, 14_train_regressor_trimesh.cpp, 24_train_regressor_ellipsoid_plane.cpp, and 34_train_regressor_ellipsoid.cpp.

stop_tol

double netdem::RegressionNet::stop_tol {1e-8}

The stopping tolerance used for determining when to stop training.

Examples

01_train_netsdf.cpp, 03_train_netspf.cpp, 04_train_regressor_trimesh_plane.cpp, 11_train_coded_netsdf.cpp, 14_train_regressor_trimesh.cpp, 24_train_regressor_ellipsoid_plane.cpp, and 34_train_regressor_ellipsoid.cpp.

---

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

• /Users/lzhshou/Documents/Research/myProjects/apaam/repo/netdem/src/mlpack/regression_net.hpp
• /Users/lzhshou/Documents/Research/myProjects/apaam/repo/netdem/src/mlpack/regression_net.cpp