What is the benefit of a positive-only sigmoid function?

[ADDA]

What is the benefit of a positive only sigmoid function? or why is it 'most often' used?

NOTES:
return value monotonically increasing most often from 0 to 1 or alternatively from −1 to 1, depending on convention; source: https://en.wikipedia.org/wiki/Sigmoid_function

 

[berkeman]

What is the benefit of a positive only sigmoid function? or why is it 'most often' used?

NOTES:
return value monotonically increasing most often from 0 to 1 or alternatively from −1 to 1, depending on convention; source: https://en.wikipedia.org/wiki/Sigmoid_function

Can you give more context to your question? Why are you asking this in the Computing forum? Is it for some modeling work you are doing? More information would make it much easier to try to answer your questions.

 

[jedishrfu]

Is this for a neural net node?

Having the activation function of each node to range from 0 (no activation) to 1 (full activation) is a common convention in neural net design.

Here are some common activation functions:

https://en.wikipedia.org/wiki/Activation_function

The sigmoid activation function provides some non-linearity to the neural net to simulate biological systems better.

 

[ADDA]

jedishrfu, you are correct. The equation, 1.0 / (1.0 + e^(-x)), comes from this video:

When I implemented a network, however, the output always converged to the error vector. Perhaps, I was wrong, I no longer have the code.

My question comes from the idea that the network has to pull down on wrong input. How can a node pull down with a positive only activation function?

 

[coscos31]

take a look at this function with only positive x values. y for x=0 equals 0

 

[coscos31]

you can slide it more to the right by increasing the nb 2 at the end of the equation but past 7 the curve is not smooth enough