Skip to content

Latest commit

 

History

History

nn

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 

Feedforward Neural Network Example

Here, a small neural network attempts to learn $f(x, y, z) = \sin(\pi x) + \sin(2y) - z$.

During backprop, gradients are computed via an approximation of the symmetric derivative $\frac{f(x+h) - f(x-h)}{2h}$.

In the code, $h$ = eps = $1\mathrm{e}{-6}$.