sigmoid function and derivative

[Terry Reedy]

On 6/26/2018 11:26 PM, Sharan Basappa wrote:
> Folks,
> 
> I know this is not a machine learning forum but I wanted to see if anyone can explain this to me.
> 
> In artificial neural network, I can understand why sigmoid is used but I see that derivative of sigmoid output function is used. I am not able to understand why.
> 
> For example:
> # convert output of sigmoid function to its derivative
> def sigmoid_output_to_derivative(output):
>      return output*(1-output)

Derivatives are used for backpropagation* of errors for adjusting 
weights.  The convenient form of sigmoid derivative is a reason to use 
sigmoids as transfer functions.

* You can look backpropagation as a method of training.

-- 
Terry Jan Reedy