grad

The C++ interface for grad. More...

void	grad (array &dcols, array &drows, const array &in)

Detailed Description

The C++ interface for grad.

grad computes the gradient of a matrix using a central difference along a given dimension. For an interior point in the matrix, \( n_i \), the central difference is computed as \( \frac{n_{i+1} - n_{i-1}}{2} \). For exterior points (those along the edge of the matrix), a one-sided difference is used.

array a; //a = [4  5  3]
         //    [8 10  9]
         //    [9 15 20]

array dx,dy;
grad(dcol,drow,a);


// Vertical gradient (along columns)
// dcol = [4   5   6]
//        [2.5 5 8.5]
//        [1   5  11]

// Horizontal gradient (along rows)
// drow = [1 -0.5 -2]
//        [2  0.5 -1]
//        [6  5.5  5]

Function Documentation

void af::grad	(	array &	dcols,
		array &	drows,
		const array &	in
	)

Parameters

[in]	in	2D
[out]	dcols	gradient along columns
[out]	drows	gradient along rows

Examples:

examples/machine_learning/neuralnetwork.cpp.