NeuralNet::Dropout Class Reference

Inheritance diagram for NeuralNet::Dropout:

Public Member Functions
	Dropout (float rate, unsigned int seed=0)
	The Dropout layer randomly sets input units to 0 with a frequency of rate at each step during training time, which helps prevent overfitting. Inputs not set to 0 are scaled up by 1 / (1 - rate) such that the sum over all inputs is unchanged.
std::string	getSlug () const override
	Dropout layer slug.
Eigen::MatrixXd	feedInputs (Eigen::MatrixXd inputs, bool training=false) override
	This method is used to feed the inputs to the layer.
Public Member Functions inherited from NeuralNet::Layer
Eigen::MatrixXd	getOutputs () const
	This method get the layer's outputs.
int	getNumNeurons () const
	This method get the number of neurons actually in the layer.
void	printOutputs ()
	Method to print layer's outputs.
virtual Eigen::MatrixXd	feedInputs (std::vector< double > inputs, bool training=false)
	This method is used to feed the inputs to the layer.
virtual void	feedInputs (std::vector< std::vector< std::vector< double > > > inputs, bool training=false)
	This method is used to feed the inputs to the layer.
const std::string	typeStr ()

Public Attributes
float	rate
float	scaleRate
unsigned int	seed
Eigen::MatrixXd	mask

Protected Member Functions
void	init (int numNeurons) override
Eigen::MatrixXd	computeOutputs (Eigen::MatrixXd inputs, bool training) override
	Drop some of the inputs randomly at the given rate.
Protected Member Functions inherited from NeuralNet::Layer
void	setOutputs (Eigen::MatrixXd outputs)
void	setOutputs (std::vector< double > outputs)
	This method is used to set the outputs of the layer.
	Layer (std::tuple< int, int > inputShape)

Friends
class	cereal::access

Additional Inherited Members
Protected Attributes inherited from NeuralNet::Layer
int	nNeurons
Eigen::MatrixXd	outputs
LayerType	type = LayerType::DEFAULT
bool	trainingOnly = false

Constructor & Destructor Documentation

Dropout()

NeuralNet::Dropout::Dropout ( float rate, unsigned int seed = 0 )

inline

The Dropout layer randomly sets input units to 0 with a frequency of rate at each step during training time, which helps prevent overfitting. Inputs not set to 0 are scaled up by 1 / (1 - rate) such that the sum over all inputs is unchanged.

Parameters

rate	Frequency of units set to 0
seed	An integer to use as a random seed

Member Function Documentation

computeOutputs()

Eigen::MatrixXd NeuralNet::Dropout::computeOutputs ( Eigen::MatrixXd inputs, bool training )

inlineoverrideprotectedvirtual

Drop some of the inputs randomly at the given rate.

Parameters

inputs

A matrix representing the inputs (features)

Returns

Inputs with some dropped values (zero-ed values) randomly

Implements NeuralNet::Layer.

feedInputs()

Eigen::MatrixXd NeuralNet::Dropout::feedInputs ( Eigen::MatrixXd inputs, bool training = false )

inlineoverridevirtual

This method is used to feed the inputs to the layer.

Parameters

inputs

An Eigen::MatrixXd representing the inputs (features)

Returns

an Eigen::MatrixXd representing the outputs of the layer

Implements NeuralNet::Layer.

getSlug()

std::string NeuralNet::Dropout::getSlug ( ) const

inlineoverridevirtual

Dropout layer slug.

Implements NeuralNet::Layer.

init()

void NeuralNet::Dropout::init ( int numNeurons )

inlineoverrideprotectedvirtual

Parameters

numNeurons

Number of neurons of the previous layers

Reimplemented from NeuralNet::Layer.

---

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

• /github/workspace/src/NeuralNet/layers/Dropout.hpp