36
$\begingroup$

$$ \begin{equation} f(x)=\frac{L}{1+e^{-k(x-x_0)}} \end{equation} $$

Fig 1. (img) Logistic Function

$$ \begin{equation} S(x)= \frac{1}{1+e^{-t}} \end{equation} $$

Fig 2. (img) Sigmoid Function

What are the differences between Logistic Function and Sigmoid Function? Is the logistic function a generalized kind of sigmoid function where you could have a higher maximum value?

$\endgroup$
2
  • $\begingroup$ $S$ is an increasing function in $t$, $f$ is increasing for values of $x \leq x_0$, decreasing for values of $x \geq x_0$ $\endgroup$
    – mic
    Mar 30, 2016 at 7:25
  • $\begingroup$ @mic I don't think that's right. If $t>0$, $S$ (the OP's "sigmoid function") is monotonically increasing. If $k>0$, (the OP's logistic function) $f$ is monotonically increasing. Both functions are monotonically increasing or decreasing, none of them increase for some values of $x$ and decrease for other values of $x$. The derivative increases then decreases, as shown here, is that what you meant by $f$? $\endgroup$ Jun 30 at 21:54

3 Answers 3

32
$\begingroup$

Yes, the sigmoid function is a special case of the Logistic function when $L=1$, $k=1$, $x_0 =0$.

If you play around with the parameters (Wolfram Alpha), you will see that

  • $L$ controls the maximum value the function can take. $e^{-k(x-x_0)}$ is always greater or equal than 0, so the maximum point is achieved when $e^{-k(x-x_0)}$ is 0 so the logistic function has value $L/1$.

  • $x_0$ controls where on the $x$ axis the growth is centered, because if you put $x_0$ in the function, $x_0 - x_0$ cancel out and $e^0 = 1$, so you end up with $f(x_0) = L/2$, the midpoint of the growth.

  • $k$ controls how steep the growth from the minimum to the maximum value is.

$\endgroup$
2
$\begingroup$

The logistic function is: $$ f(x) = \frac{K}{1+Ce^{-rx}} $$ where $C$ is the constant from integration, $r$ is the proportionality constant, and $K$ is the threshold limit.

Assuming the limits are between $0$ and $1$, we get $\frac{1}{1+e^{-x}}$ which is the sigmoid function.

$\endgroup$
2
  • $\begingroup$ what is $C$ -- I don't see it in the OP's question. What does "constant from integration" mean? Do you have an example to cite, so we can learn more? $\endgroup$ Jun 30 at 22:00
  • $\begingroup$ Trying to answer my question, this resource, ("A First Course on Time Series Analysis") says: "the function $\frac{β3}{1 + β2 exp(−β1t)}$ with $β1, β2, β3 ∈ \!R $ **is the widely used logistic function** Its $β2$ is same as your $C$. This video is also consistent with your definition. But I'm still trying to understand the significance of $C$, aka $β2$, and why Wikipedia doesn't mention it. $\endgroup$ Jul 6 at 23:20
2
$\begingroup$

I would like to say in opposite way to the answer "the sigmoid function is a special case of the Logistic function" into "The Logisitic function is a special case of the sigmoid function". All S shape curved monotonically increasing fuction being confined a and b are sigmoid functions.

$\endgroup$
2
  • 1
    $\begingroup$ That is also how Wikipedia describes the sigmoid functions. $\endgroup$
    – Jiho Noh
    Mar 21 at 19:08
  • $\begingroup$ Interesting - I upvoted you -- Wikipedia does say that sigmoid functions is a general category of any "S-shaped curve", and gives logistic function as just one example. However, Wikipedia labels the "logistic function" example with the definition $f(x) = \frac{1}{1+e^{-x}}$ , which is what the OP calls the "sigmoid function". There are more parameters in what the OP calls the "logistic function". So you're right, but I also think the accepted answer is right in addressing the OP's definitions. $\endgroup$ Jun 30 at 21:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.