input variables
and produces
outputs, corresponding to the cross-multiplication of the inputs:
where
,
.
Example:
The resulting terms, combinend with a linear transform, can model quadric decision boundaries (parabola, hyperbola, ellipsoids and spheres), which are richer in structure than classic linear discriminants.
Note:
Functions | |
| int | gnn_quadric_f (gnn_node *node, const gsl_vector *x, const gsl_vector *w, gsl_vector *y) |
| Computes the output. | |
| int | gnn_quadric_dx (gnn_node *node, const gsl_vector *x, const gsl_vector *w, const gsl_vector *dy, gsl_vector *dx) |
Computes
. | |
| gnn_node * | gnn_quadric_new (int input_size) |
| Creates a Quadric Discriminant Transfer function node. | |
|
||||||||||||||||||||||||
|
This functions computes the gradient of the Quadric Discriminant activation function, given dy (
Definition at line 146 of file gnn_quadric.c. |
|
||||||||||||||||||||
|
This functions evaluates the Quadric Discriminant Transfer function.
Definition at line 92 of file gnn_quadric.c. |
|
|
This function creates a node of the gnn_quadric type. This node produces the second-order terms of its inputs. For a complete review, see gnn_quadric : Quadric Discriminant Activation Function..
Definition at line 208 of file gnn_quadric.c. |
1.2.18