---

gnn_tanh : Hyperbolic Tangent Activation Function.
[Atomic Nodes.]

---

Detailed Description

This node type implements the Hyperbolic Tangent activation function given by:

where the index runs over all inputs / outputs.

This function is a member of the more general class of sigmoid functions, which have the property to "squash" its inputs to a very restricted output interval.

The Hyperbolic Tangent is symmetric and its outputs lie between . It has been said that it leads to faster convergence in training than the Logistic Sigmoid.

Functions
int	gnn_tanh_f (gnn_node *node, const gsl_vector *x, const gsl_vector *w, gsl_vector *y)
	Computes the output.
int	gnn_tanh_dx (gnn_node *node, const gsl_vector *x, const gsl_vector *w, const gsl_vector *dy, gsl_vector *dx)
	Computes .
gnn_node *	gnn_tanh_new (int input_size, double a, double b)
	Creates a Hyperbolic Tangent Activation Function node.
gnn_node *	gnn_tanh_standard_new (int input_size)
	Creates a standard Hyperbolic Tangent function node.
gnn_node *	gnn_tanh_enhanced_new (int input_size)
	Creates a Hyperbolic Tangent function node.
double	gnn_tanh_get_a (gnn_node *node)
	Returns the amplitude constant .
void	gnn_tanh_set_a (gnn_node *node, double a)
	Sets the amplitude constant .
double	gnn_tanh_get_b (gnn_node *node)
	Gets the streching factor .
void	gnn_tanh_set_b (gnn_node *node, double b)
	Sets the streching factor .

---

Function Documentation

int gnn_tanh_dx (  gnn_node *    node, const gsl_vector *    x, const gsl_vector *    w, const gsl_vector *    dy, gsl_vector *    dx )  [static]

This functions computes the gradient of the hyperbolic tangent function, given dedy ( ). The function is, Parameters: node  A pointer to the gnn_tanh node. x  The input vector . w  The parameter vector (which is empty in this case). dy  The vector . dx  The output buffer vector . Returns: Returns 0 on success. Definition at line 148 of file gnn_tanh.c.

gnn_node* gnn_tanh_enhanced_new (  int    input_size )

This function creates a node of the gnn_tanh type. where . The constants are chosen so that when the layer is feed with a linear combination layer with normalized outputs its own outputs have an effective gain of 1. Parameters: input_size  The input size . Returns: A pointer to a new gnn_tanh node. Definition at line 281 of file gnn_tanh.c.

int gnn_tanh_f (  gnn_node *    node, const gsl_vector *    x, const gsl_vector *    w, gsl_vector *    y )  [static]

This functions evaluates the hyperbolic tangent function. Parameters: node  A pointer to the gnn_tanh node. x  The input vector . w  The parameter vector (which is empty in this case). y  The output buffer vector . Returns: Returns 0 on success. Definition at line 101 of file gnn_tanh.c.

double gnn_tanh_get_a (  gnn_node *    node )

Parameters: node  A pointer to the gnn_tanh node. Returns: The amplitude of the sigmoid. Definition at line 294 of file gnn_tanh.c.

double gnn_tanh_get_b (  gnn_node *    node )

Parameters: node  A pointer to the gnn_tanh node. Returns: The streching factor . Definition at line 320 of file gnn_tanh.c.

gnn_node* gnn_tanh_new (  int    input_size, double    a, double    b )

This function creates a node of the gnn_tanh type. This node computes where , and constant parameters. Parameters: input_size  The input size . a  The factor. b  The factor. Returns: A pointer to a new gnn_tanh node. Definition at line 204 of file gnn_tanh.c.

void gnn_tanh_set_a (  gnn_node *    node, double    a )

Parameters: node  A pointer to the gnn_tanh node. a  The amplitude factor. Definition at line 307 of file gnn_tanh.c.

void gnn_tanh_set_b (  gnn_node *    node, double    b )

Parameters: node  A pointer to the gnn_tanh node. b  The new streching factor . Definition at line 333 of file gnn_tanh.c.

gnn_node* gnn_tanh_standard_new (  int    input_size )

This function creates a layer of the gnn_tanh type: where . Parameters: input_size  The input size . Returns: A pointer to a new gnn_tanh node. Definition at line 262 of file gnn_tanh.c.

---