1
$\begingroup$

I am not familiar with ML background. Thus have this simple question. Suppose I have training data for my dependent variable which depends on say 3 independent variables. Suppose I use some regression method and perform a fit according to that model on my training data.

Suppose now I pass independent variables from testing data and see the predicted value of my dependent variable.

Now, my question is, when I perform this non-parametric fit, what type of function gets stored or what actually happens such now when I supply some testing data, it predicts almost actual value.

Like, if I say, I perform a polynomial fit. I know that it is a line y = mx + c something and thus on giving x, I can predict y. And here the value of m and c gets derived from the training data. So now what type of function or parameters are involved in non-parametric fit?

Also, if I use scikit-learn, is there a method to see the values of these parameters derived from training data fit?

Improve this question
$\endgroup$
4
  • $\begingroup$ What does the documentation say for the specific scikit-learn functions you intend to use? $\endgroup$
    – rickhg12hs
    Jun 5, 2018 at 6:54
  • $\begingroup$ Well, this was a general question. However, I am planning to use quantile regression in specific and all that it mentions is that you perform a non parametric fit to predict the value using predict(X) function... $\endgroup$
    – kg__
    Jun 5, 2018 at 6:58
  • $\begingroup$ Also, what I understood is that, for quantile regression, by performing a fit, we actually determine the so-called quantile distribution function, if I am not mistaken? $\endgroup$
    – kg__
    Jun 5, 2018 at 8:57
  • $\begingroup$ What is stored will depend on what kind of nls fit you are using. $\endgroup$
    – Peter Flom
    Jun 5, 2018 at 18:24

0

Reset to default

Your Answer

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