[HN Gopher] Cross-Validation FAQ (2022)
___________________________________________________________________
Cross-Validation FAQ (2022)
Author : Tomte
Score : 34 points
Date : 2023-07-30 14:03 UTC (8 hours ago)
(HTM) web link (avehtari.github.io)
(TXT) w3m dump (avehtari.github.io)
| jwilber wrote:
| And here's a shorter, visual explainer on cross-validation:
|
| https://mlu-explain.github.io/cross-validation/
|
| What's really interesting about k-fold cv is that most textbooks
| claim that the greater the value of k in k-fold cross-validation,
| the more variance our estimates tend to have, with Leave-One-Out
| Cross-Validation (LOOCV) exhibiting exceptionally high variance.
| This is frequently rationalized by asserting that a larger k will
| lead to models being trained on very similar data, thereby
| creating essentially identical models. Nevertheless, this
| rationale does not always hold true and can be significantly
| influenced by the specifics of the model and dataset in use!
___________________________________________________________________
(page generated 2023-07-30 23:01 UTC)