Talk:Conformal prediction
![]() | This article is rated Start-class on Wikipedia's content assessment scale. |
There is a template saying "Please improve this by adding secondary or tertiary sources", and these are some possible secondary sources (in addition to those already mentioned in the article):
- An excellent review "Conformal Prediction: a Unified Review of Theory and New Challenges" by Matteo Fontana, Gianluca Zeni, and Simone Vantini, to appear in "Bernoulli" (it can be downloaded from the journal's web page, and also available on arXiv).
- The review "Methods to Compute Prediction Intervals: A Review and New Results" by Qinglong Tian, Daniel J. Nordman, and William Q. Meeker (to appear in "Statistical Science" and available from the journal's web site) is not just about conformal prediction (which is covered in Section 9.2), but it embeds conformal prediction into a more general picture of predictive inference.
- The nice review "A Gentle Introduction to Conformal Prediction and Distribution-Free Uncertainty Quantification" has only been published as an arXiv report (https://arxiv.org/abs/2107.07511) so far.
- There is also an older book, "Conformal Prediction for Reliable Machine Learning (edited by Balasubramanian et al., 2014), which contains reviews by different authors.
Ukookami (talk) 18:30, 4 December 2021 (UTC)
Compare to Bootstrap prediction intervals
[edit]Would it make sense to compare conformal prediction with prediction intervals obtained from bootstrapping? Biggerj1 (talk) 11:57, 30 September 2023 (UTC)
Illustration like in the YouTube video needed!
[edit]Biggerj1 (talk) 19:12, 1 October 2023 (UTC)
- @Dicklyon: what do you think, could you make a picture or animation similar to the YouTube video? Biggerj1 (talk) 10:59, 24 October 2023 (UTC)
- Interesting. Not a topic I'm familiar with, but I see a sort of consistent illustration scheme out there. Maybe later... Dicklyon (talk) 20:07, 24 October 2023 (UTC)
Which uncertainty is measured?
[edit]Does conformal prediction measure both aleatoric and epistemic uncertainty? How will the predicted intervals vary when training data become sparse? A bootstrap method would indicate a bigger uncertainty in these regions. It appears to me that conformal prediction would not do so. Can somebody give a literature hint or provide an answer in the article? Biggerj1 (talk) 18:47, 2 October 2023 (UTC)