0
$\begingroup$

Let's say we're going to train a classifier with the full data set. There's also a reject logic for ambiguous regions in the data. So, at the end, the final system outputs reject or 0 or 1. That is, reject data points in regions with high ambiguity, otherwise use the original model predictions.

  1. If you were to see a scatterplot with a boxplot accompanying it, how would you interpret the data in this plot?
  2. How would you reject ambiguous data points?

enter image description here

enter image description here

Improve this question
$\endgroup$
2
  • 1
    $\begingroup$ How about setup as a binary classification problem, and in addition to the 0/1 label, also output a (calibrated) confidence of the prediction, then 'reject' those with low confidence? $\endgroup$
    – lpounng
    May 8 at 4:29
  • 1
    $\begingroup$ FYR see this post $\endgroup$
    – lpounng
    May 8 at 4:32

0

Reset to default

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.