-1

I did some search by google and got a few answers but they all have their own problems. For example

  1. The result of sigmod will between 0 and 1.

This is good but sigmod function is not the only function whose result is between 0 and 1

  1. It is monotonic function but again there are a lot of monotonic function.

So why sigmod? Some how I think it has some connection with central limit theorem although I do not know what it is.

So can you explain why sigmod in logic regression? Why it is good to solve classification problems?

Kramer Li
  • 2,284
  • 5
  • 27
  • 55

1 Answers1

0

I suggest to move this question to CrossValidated since that's a more data science / statistics / math focused resource.

One of the main reasons you want to have a function between 0 and 1 and monotonic ascending is because that way you can transform 'scores' into 'probabilities'. Namely a probability must be non negative and its Distribution (cumulative) must be monotonic.

You are right, there are more than one function that has this requisites but this function is also differentiable and continuous which makes it convenient for probabilistic modeling. Plus it has several mathematical properties that involve refactoring and derivative in the functions own term.

Diego Aguado
  • 1,604
  • 18
  • 36