Holds all information related to regularization terms in an error function. More...

Public Member Functions
	Regularization (double l1Penalty=0.0, double l2Penalty=0.0, double maxSquaredWeightNorm=0.0)

Public Attributes
double	l1Penalty
	Penalty for absolute weights, encourages sparse weight matrices. More...

double	l2Penalty
	Penalty for squared weights. More...

double	maxSquaredWeightNorm
	Maximum value for squared norm of the weight vector of a single neuron. More...

Detailed Description

Holds all information related to regularization terms in an error function.

Usually we do not penalize biases.

Constructor & Destructor Documentation

OpenANN::Regularization::Regularization	(	double	l1Penalty = `0.0`,
		double	l2Penalty = `0.0`,
		double	maxSquaredWeightNorm = `0.0`
	)

inline

double OpenANN::Regularization::l1Penalty

Penalty for absolute weights, encourages sparse weight matrices.

double OpenANN::Regularization::l2Penalty

Penalty for squared weights.

double OpenANN::Regularization::maxSquaredWeightNorm

Maximum value for squared norm of the weight vector of a single neuron.

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