0
$\begingroup$

Is there a way to weight instances based upon the values of a feature? For instance for making a sample better resemble a population on certain parameters.

$\endgroup$
1
  • $\begingroup$ Are these weights static? Or it may change along with the training? $\endgroup$
    – Icyblade
    Commented Feb 26, 2017 at 1:45

2 Answers 2

1
$\begingroup$

Try reading up on Loss/ Cost functions. Samples that are closer to your model will be assigned a smaller cost while others a higher one.

https://en.wikipedia.org/wiki/Loss_function

$\endgroup$
1
$\begingroup$

In Orange documentation - Loading your Data, it says to prefix your column with w# to be marked as instance weights column.

$\endgroup$

Your Answer

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

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