I have a question that is about data mining (Which may relate to cross validated), but it's mostly mathematical.

The question is:


How to fit a function with multi-parameters to data with noise?


I will give example: My data is: $g(1),g(2),g(3) = 2.1,3.95,8.15$ (Previously being $2$, $4$, $8$)
The function I am trying to fit in is: $h(x) = a^{bx}$. The fit function: $d(x) = (h(x) - g(x))^2$ So to calculate to overall error with my parameters I just do: $f(a,b) = d(1) + d(2) + d(3)$

To minimize the overall error I thought just to find $f'(a,b)$, and in this way find when $f(a,b)$ is the closest to $0$, but I don't know. How to find the deviation of a function with multi-parameters? Is there a workaround to this?

?? What should I put here?

Does it fit in math SE?

  • $\begingroup$ I think it would fit, but I would still ask it on Cross Validated instead (Disclaimer: I do not know that site well.) $\endgroup$
    – quid Mod
    Jun 26, 2015 at 12:10
  • $\begingroup$ You asked what you should put as tags. A natural choice seems to be (data-mining). Notice that stats.SE also has a tag called (data-mining). (There are much more questions with this tag on stats.SE than on math.SE.) Maybe looking on some questions having this tag on both sites might be useful before asking. (It might help you with the decision where to ask. And you might even find some closely related questions.) $\endgroup$ Jun 26, 2015 at 12:15

1 Answer 1


There is no reason that you couldn't post this on Math.SE if you were to decide that the community here would have more appropriate expertise.

If you choose to do so, then you would do best to cut out extraneous contextual details and present the problem as a purely mathematical one.

If, however, you're looking for suggestions to improve your algorithm in the context of data mining and feel that the other SE site has more to offer, then post it there. You could even post to both sites if you present the question appropriately to each one.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .