Activation function for multilayer neural network?

Question

Neural networks

I solve the xor problem for a multilayer neural network with back propagation of an error. The output is two numbers a and b [equal to 0 or 1 ] and the input should get xor ( a , b ) I
use the sigmoid function. 1 / ( 1 + exp ( - x ) ) of
course I can't get an exact answer, only some kind of approximation.
Are there any functions that can exactly solve this problem?
If there is, then you can have an example of a function.

Answer 1

In your case, I think it would make sense to round to the nearest integer. But if you want to get a more accurate result, then there are a huge number of functions that you can try as an activation function for your neurons. In particular:
Linear transfer function
Threshold transfer function
Sigmoid transfer function
Logistic function (which you are currently using)
Hyperbolic tangent
Modified hyperbolic tangent
Radial basis transfer function
And many, many more functions